Beetle species responses to tropical forest fragmentation.
Deforestation and forest fragmentation are clearly causing a loss of species from tropical forests (Saunders et al. 1991, Andren 1994, Vitousek 1994, Didham et al. 1996, Turner 1996, Turner and Corlett 1996, Laurance and Bierregaard 1997). Simplistically, we know that some invertebrate species are affected by forest fragmentation; we suspect that many, if not most, species are probably affected, either directly or indirectly, by fragmentation; and we know that not all species will be adversely affected. A first step toward ascertaining the proportion of species affected by forest fragmentation, and hence the magnitude of biodiversity loss (Myers 1986, Wilson 1988, Sayer and Whitmore 1991, Whitmore and Sayer 1992, Barbault and Sastrapradja 1995, Lawton and May 1995), must be an analysis of species responses to fragmentation in a diverse, tropical invertebrate assemblage.
At present, we have a qualitative knowledge of the types of responses exhibited by invertebrate species to habitat fragmentation, but the patterns are strikingly idiosyncratic. Essentially we have no clear consensus as to why some species are more susceptible to fragmentation than others. There are many explanations for the lack of any kind of framework for interpreting invertebrate responses. The first is simply a dearth of data from a wide range of species. The species that have been studied are perhaps also atypical of the majority of invertebrates because they are large (e.g., butterflies: Lovejoy et al. 1986, Brown and Hutchings 1997; dung beetles: Klein 1989), functionally unique (e.g., leaf-cutter ants: Vasconcelos 1988), or specialized (e.g., euglossine bees: Powell and Powell 1987, Becker et al. 1991).
Second, even though the compilation of data on invertebrate responses to fragmentation lags decades behind that for vertebrate species, it is still surprising that little attempt has been made to contrast the attributes of invertebrate species that confer susceptibility or resilience to fragmentation. Correspondingly, generalities that have been made about where, why, and how vertebrate species are affected by fragmentation (e.g., Soule 1983, Pimm et al. 1988, Laurance 1991, Saunders et al. 1991, Tracy and George 1992, Andren 1994, Gaston 1994) cannot yet be made for invertebrates. Such generalities are that large-bodied, rare, poorly dispersing species, particularly in higher trophic levels, are most prone to extinction due to habitat destruction (Leck 1979, Soule 1983, Diamond 1984, Pimm et al. 1988, Burbidge and McKenzie 1989, Laurance 1991, Tracy and George 1992, Gaston 1994, Tilman et al. 1994, Lawton 1995, Holt 1996). We do not know if these same generalities are also applicable to invertebrate species. Of course, even for vertebrates the data are often conflicting. For example, body size has been found to have positive, negative, or no effect on the risk of species' extinctions (Diamond 1984, Pimm et al. 1988, Maurer et al. 1991, Tracy and George 1992, Gaston 1994), and in various studies rarity has either been found to make species more susceptible to habitat destruction (review in Gaston 1994), or conversely less susceptible, because of a trade-off between competitive dominance and dispersal abilities that makes abundant species less likely to persist in a fragmented landscape (Nee and May 1992, Tilman et al. 1994).
Third, a notable deficiency of invertebrate and vertebrate habitat fragmentation studies alike is the resolution of interacting variables, such as habitat loss, fragment area, distance from forest edge, degree of spatio-temporal isolation, fragment shape, and habitat matrix effects (Kareiva 1990, Lord and Norton 1990, Laurance and Yensen 1991, Saunders et al. 1991, Doak et al. 1992, Murcia 1995, Didham 1997a). Studies typically draw samples from the centers of forest fragments as an indicator of an overall fragmentation effect, rather than explicitly controlling for confounding variables. "Fragment area effects" are usually the proffered explanation for variation in population densities or species richness, but individual species could equally be responding to edge effects, variation in rates of dispersal between fragments, or any number of variables in combination. Any perceived pattern of response across species is likely, therefore, to appear idiosyncratic, as the examples below serve to point out.
The general flavor of invertebrate species responses to fragmentation is captured by Margules et al. (1994), expounding the more general proposition of Robinson et al. (1992), that different organisms respond differently to the same level of habitat fragmentation. Many invertebrate species are recognized as deep-forest specialists (e.g., Main 1987, Powell and Powell 1987, Vasconcelos 1988, Harper 1989, Klein 1989, Margules et al. 1994), while other species apparently prefer small fragments. Interestingly, empirical examples of small-area preference tend to be mediated by interspecific interactions, but this may be purely artifactual (lack of data, once again). For example, Apis mellifera (Hymenoptera, Apidae) attained higher densities in small rather than large forest fragments in Argentina where they replaced the decimated native insect pollinator fauna (Aizen and Feinsinger 1994); the inference is that specialist species are negatively affected by fragmentation, while generalist species benefit (see also Kitahara and Fujii 1994). Similar cases have been noted for closely related, competing beetle species feeding on the same host plant (Bach 1988), and bumble bee species feeding at the same flowers (Sowig 1989), in habitat patches of different sizes.
Many more invertebrate species are known to be edge specialists and positively affected by fragmentation (e.g., Bellinger et al. 1989, Cusson et al. 1990, McCann and Harman 1990, Cappuccino and Root 1992, Clopton and Gold 1993, Roland 1993, Roland et al. 1997). Naturally, there are other species that show no edge pattern, or are edge avoiders (Bedford and Usher 1994). Because the proportion of edge habitat increases with decreasing fragment area, however, edge specialists will also appear to prefer small fragments and edge avoiders to prefer large fragments. Thus it becomes impossible to separate environmentally based edge effects from demographic changes due to reduced fragment area, or direct vs. indirect species interactions. The same can be said for many other interacting variables.
Are these patterns, then, governed by rules? Species attributes that may explain significant invertebrate species responses to fragmentation have not been considered. And what of the species that are not affected by fragmentation? The most neglected patterns of species responses to fragmentation are the nonsignificant ones. Determining species traits that confer resilience to fragmentation may be as profitable as assessing traits that predispose species to local population decline.
Individual species responses are the key to understanding the mechanistic bases for fragmentation-induced changes in biological communities. Studies of the biological impact of forest fragmentation do not often take this into account. In this paper, we analyze beetle species responses to fragmentation in an experimentally fragmented tropical forest landscape in Central Amazonia. From this analysis we develop a first empirical classification of invertebrate population density responses to the two most important fragmentation processes, fragment area effects and edge effects (in a form of incomplete factorial design), while holding other variables constant. This process is just the first step toward ascertaining rates of species loss from forest fragments and hence local and regional loss of biodiversity. We then compare and contrast species traits (body size, trophic group, rarity, and population variability) across many species within the same assemblage and relate these to trends in population density and probability of local extinction from tropical forest fragments.
The study was carried out from January to August 1994 at the Biological Dynamics of Forest Fragments Project (BDFFP), 80 km north of Manaus, Central Amazonia, Brazil (2 [degrees] 25' S, 59 [degrees] 50' W). The BDFFP is administered by the Instituto Nacional de Pesquisas da Amazonia (INPA) and the Smithsonian Institution, and is the only experimental tropical forest fragmentation project of its kind in the world, offering unique opportunities to study the impact of fragmentation on biotic and abiotic processes in tropical forest fragments of known age and history.
The forest is a uniform upland dry (terra firme) forest on yellow alic latosol soils of high clay content (Chauvel 1983, Lovejoy and Bierregaard 1990, Chauvel et al. 1991, Camargo 1992, Camargo and Kapos 1995). Vegetation of the region is described by Prance (1990), and of the BDFFP fragments by Rankin-de-Merona et al. (1992). Forest disturbance in the area is principally due to cattle ranching, with pastures created and maintained by fire (Jordan 1986, Fearnside 1989, 1990, Nepstad et al. 1993, 1996). For further description of the study area see Lovejoy et al. (1986), Bierregaard et al. (1992), and Bierregaard and Stouffer (1997).
The leaf-litter beetle fauna was sampled with respect to fragment area and distance from forest edge. Forest fragments and continuous forest edges were selected to minimize unexplained variation in community composition. Hence, all sites were upland (terra firme) forests on similar soil types, forest edges were adjacent to well-maintained pasture without secondary regrowth, all but one of the edges were west-facing (see below), the shape of all fragments was similar, and the year and distance of isolation from continuous forest were fairly constant across sites (Table 1).
TABLE 1. Full description of BDFFP sites sampled in this study (modified from Lovejoy et al. 1986). BDFFP reserve Size Code in Location number (ha) this study (farm) 2107 1 1-ha 1 Dimona 2108 1 1-ha 2 Dimona 3209 10 10-ha 1 Porto Alegre 1202 10 10-ha 2 Esteio 2303 100 100-ha 1 Dimona 3304 100 100-ha 2 Porto Alegre 1401 CF([dagger]) Edge 1 and Esteio Edge 2 1501 CF([dagger]) Interior 1 and Esteio Interior 2 BDFFP reserve number Year of isolation Description 2107 1984 Gully leaves reserve on the west. Unusual patch of understory bamboo. 2108 1984 Gentle topographic relief sloping down to northwest, includes dense stand of bamboo in southwest corner. 3209 1983 Dry, gentle relief. One poorly drained area devoid of understory palms. 1202 1980 No stream, plateau at west, border slopes steeply to east, back up to hill crest 100 m in from north border. One ha in southeast corner burned by farm grass fire, now regenerating. 2303 north and west edges Rolling bisected terrain, high hill 1984, completed in north-west, drainage with 1990 gullies to southeast. Very steep gully leaves northeast corner. Swampy area along south border. Soil sandy, canopy tree density high. Several hectares in northwest with poor drainage and a hummock effect. 3304 1983 Undulating terrain, triple-headed stream system along east half. Unusual Moraceae stand in northwest. Also some Buriti stands in poorly drained areas. 1401 edges created 1980 Many streams, accentuated local relief, 7-10 streams drain into a large stream 2 m wide draining to northeast. Soils in south less sandy than in north. Large swampy areas (respectively) in northeast. 1501 n/a Not available, Edge Nearest BDFFP type Dominant continuous reserve (open/ vegetation Aspect forest number closed) at edge of edge (m) 2107 closed Cecropia west ~200 2108 closed Cecropia west ~200 3209 closed Cecropia west ~500 1202 closed Cecropia west ~300 2303 closed Cecropia west ~100 3304 open Vismia north ~200 1401 closed Cecropia and west (both) n/a and Vismia open 1501 n/a n/a n/a n/a BDFFP reserve number Soil description 2107 88% clay and very clayey, 12% clayey 2108 loam, sandy clay loam and sandy loam 3209 1202 100% clay and very clayey 2303 81% clay and very clayey, 19% sandy through sandy clay (45-60% sand) 3304 ? 1401 ? 1501 ? Note: For explanation of open/closed edge types, see Didham and Lawton (in press). Soil data are from Rankin-de-Merona et al. (1992), except reserve 2303 (Camargo 1992). Where no soil data are available, the average values given by Rankin-de-Merona et al. (1992) are a good approximation: 78% clay and very clayey, 18% clayey loam, sandy clay loam and sandy clay, and 4% sandy, loamy sand, and sandy loam. "Clay and very clayey" indicates >40% clay and <50% sand, while "sandy" indicates >50% sand. n/a = not applicable. ([dagger]) CF = continuous forest.
The sampling design was based on a comparison of two independent transects sampled at each of three locations: (A) deep within undisturbed continuous forest ([is greater than] 10 km from the nearest edge); (B) from the edge to the interior of continuous forest; and (C) from the edge to the interior of two 100-ha isolated forest fragments. The two 100-ha fragments (BDFFP numbers 2303 and 3304, designated 100-ha 1 and 2, respectively) were located ~14 km apart; the two continuous forest edges, Edge 1 and Edge 2, were separated by a distance of 2 km (both on the western edge of continuous forest 1401); and the deep-forest control plots, Interior 1 and Interior 2, were 2 km apart (both in control site 1501). All forest edges abutted well-maintained pasture and were west-facing, with the exception of 100-ha 2 where the only edge abutting well-maintained pasture was north-facing.
Beetles were collected at seven distances along each of the six transects: 0, 13, 26, 52, 105, 210, and 420 m. This sampling protocol reflected the a priori expectation that the rate of change in beetle community structure would be greatest near the forest edge. In addition, to assess beetle communities in small forest fragments, two 10-ha fragments (numbers 3209 and 1202, designated 10-ha 1 and 2, respectively) were sampled at 105 m from the edge, and two 1-ha fragments (numbers 2107 and 2108, designated 1-ha 1 and 2, respectively) were sampled at 52 m from the edge.
Sampling leaf-litter beetles
Twenty random, 1-[m.sup.2] leaf-litter samples were collected at each of the 46 sites over a 5-mo period. Within logistical constraints of movement throughout the study area, daily sampling was randomly allocated between different transects and sites to prevent bias arising from daily and seasonal variation in activity patterns of beetles. Typically, 4-6 [m.sup.2] of leaf litter were collected from any one site on each of four visits spread approximately evenly throughout the sampling period. All friable leaf litter was scraped rapidly from the quadrat and placed in a large bag-sieve to minimize beetle escape. The material was immediately sieved over a 9mm mesh by vigorously shaking the bag-sieve for ~5 min. Beetles with a cross-sectional diameter larger than 9 mm are rare in tropical forest leaf litter, and none was ever observed in discarded leaf material. The fine, sieved litter containing beetles was then transported to the laboratory in individual cotton bags. Beetles were extracted using the Winkler method (Besuchet et al. 1987), whereby sieved leaf litter was carefully placed into coarse mesh bags, which were then suspended gently inside a large sealed cloth bag and hung for 3 d. As the leaf litter dried out, beetles sensitive to desiccation moved downwards through the mesh bag and fell into a jar of alcohol below. We operated 40 Winkler bags continuously for 5 mo.
The Winkler method is sensitive to climate and collection methods and hence requires a strict, standardized methodology. We only collected samples from plateau forest areas (i.e., transects were not located in gullies or seasonally flooded areas), and only on dry mornings when there had been no rain the previous late afternoon or night. Leaf-litter sampling was discontinued if it rained. All samples were dried for 3 d, and no extra hand-sorting of litter was performed. Despite these restrictions, the Winkler method is still inherently a "relative" trapping method and, as with most invertebrate sampling methods, does not sample all taxa with equal efficiency. However, it is a particularly good method for the rapid and efficient extraction of beetles from large numbers of samples (Besuchet et al. 1987, Nadkarni and Longino 1990).
Beetles were sorted to family (Appendix A) and morphospecies (Hammond 1994), hereafter referred to as species. All taxa were sorted by the authors with the aid of The Natural History Museum, London, collections. In problematic cases, specimens were dissected for genitalial characters or checked by specialists (see Acknowledgments). Nomenclature of beetle families and subfamilies follows Lawrence and Newton (1995). Specimens are deposited at the Department of Entomology, INPA, Manaus, Brazil, and The Natural History Museum, London, UK.
The error rate in species sorting was estimated by a comparison of species sorting of Pselaphinae (Staphylinidae) by the authors with a check by a specialist in this group. Of 109 species of Pselaphinae recognized, one species was considered by the specialist to be incorrectly assigned, an error rate of [is less than] 1%. The Pselaphinae are among the most diverse and poorly known groups of beetles in tropical leaf litter, here represented by 109 species in 42 genera, of which at least 7 genera were undescribed. The proportion of undescribed species in the total assemblage is estimated to be ~90%.
Beetle species were assigned to six trophic groups according to Hammond (1990) (Appendix A): fungivores, herbivores, predators, saprophages, xylophages, and xylomycetophages (specialists on "ambrosia" fungi inside wood). There were no positive identifications of parasitoids in the assemblage, so this trophic group was not included. Where only one feeding biology was known for a family, all species were assigned to that trophic group. In other cases, where multiple feeding biologies were known to occur, species were assigned on an individual basis using mouthpart and general morphological characters, as well as published details of the feeding biology of the genus, or of related genera.
Measurement of environmental variables
Fragment area, distance from forest edge, spatial location within the study area, and nine measures of environmental variation between sites were taken (Appendix B) (see also Didham and Lawton, in press).
Microclimate.--From mid-July to mid-August 1994 three consecutive sets of data on air temperature (TEMP) and evaporative drying rates (EVAP) were obtained between 1000 and 1500 by continuously walking back and forth along the transects, taking readings at each site. These data characterize early to mid-dry-season daytime gradients.
Air temperature ([degrees] C) was measured 1.8 m above ground level with an electronic temperature probe. Air temperatures are expressed as the difference between observed air temperature at edge sites and "expected" continuous forest air temperature at the same time of day, using a standardized air temperature curve calculated empirically for continuous forest (n = 740). Temperature differentials for the seven sites along each of the four edge transects are the untransformed means of n = 33 (site 100-ha 1), n = 27 (100-ha 2), n = 23 (Edge 1) or n = 30 (Edge 2) measurements. Site means for continuous forest are the untransformed mean residuals around the standardized curve (Appendix B).
Evaporative drying rate was measured using a simple experimental apparatus consisting of a test tube and filter paper wick (Didham and Lawton, in press). Rate of water loss (in milliliters per hour) at ground level was calculated using five tubes located randomly at each site. Approximately hourly measurements were taken for three consecutive days, in conjunction with air temperature measurements. Evaporative drying rates are the untransformed means of the three daily rates (Appendix B).
Air temperature and evaporative drying rates were not measured in 10-ha or 1-ha fragments, so values were estimated from the average of 100-ha sites at equivalent distances from the forest edge (Appendix B).
Vegetation structure.--At each site, 10 random 5 x 5 m quadrats were sampled for canopy height (CANHEIGHT) and density (CANDENS), using modifications of the methods of Hubbell and Foster (1988). To reduce sampling error, five measurements of canopy height and density were taken in each of the 10 5 x 5 m quadrats at each site, and the resulting mean values were used in subsequent analyses.
Canopy height was estimated by one observer (R. K. Didham) against the height of a 3-m sighting pole; this method underestimated actual canopy height by ~3-5 m, when compared with measures made with a range finder at the same sites (Camargo 1992). Means are back-transformed from cubic-transformed variates, except for 100-ha 1 sites where median values are used because data could not be normalized (Appendix B).
A 3-m pole was used as a vertical sighting instrument to estimate the foliage density at three height intervals, 0-2 m, 2-5 m, and [is greater than] 5 m. Data are presented here for the canopy ([is greater than] 5 m) category. Foliage density was scored on an arbitrary scale of 0 to 3 (0 = no foliage intercepting the line of sight, 1 = trace-33% foliage in line of sight, 2 = 33-66% foliage in line of sight, and 3 = 66-100% foliage in line of sight). The data were not normally distributed and median values are presented (Appendix B).
Litter structure.--Three leaf-litter variables were measured from January to May 1994: litter depth (LIT-DEPTH), biomass (LITBIOM), and moisture content (LITMOIST). Twenty 25 x 25 cm quadrats were randomly located at each site and all fine litter (i.e., excluding woody debris [is greater than] 2 cm diameter) was collected down to the compact soil layer. The litter was weighed, and then oven-dried and weighed again; moisture content (%) was calculated as (1 - dry mass/wet mass). Five litter depth measurements (in millimeters) were taken immediately adjacent to the litter quadrats and the mean of these used as a single variate in subsequent analyses. Mean litter depths are untransformed, mean litter biomass values are back-transformed from In-transformed variates, and mean litter moisture contents are back-transformed from arcsine square-root transformed percentages (Appendix B).
Ground cover estimates were made in conjunction with the 5 x 5 m vegetation quadrats. Percent ground cover was scored in 10 categories. Median values for the two predominant categories, percent leaf-litter cover, and percent twig cover are analyzed here (Appendix B).
Distance between sites.--A preliminary analysis (Didham 1997b) suggested the possibility of significant variation in beetle species composition due to site location within the study area. All sites sampled were ostensibly of the same habitat type, but it is possible that there were undetected underlying environmental gradients across the study area. Thus, there may be intrinsic species turnover between sites due to beta diversity (Whittaker 1970, Rosenzweig 1995). To test for species turnover across the study area, sites were scored for approximate ground distance along the major east-to-west axis (Appendix B).
Species richness and composition.--Species accumulation curves were calculated for undisturbed continuous forest and for the entire sample set, using a BASIC computer program. For undisturbed continuous forest, species accumulation with increasing sample size was calculated from 100 replicate random draws of each of (1, 6, 11, 16, ..., 280) 1-[m.sup.2] samples, drawn without replacement from a total of 280 samples. Similarly, for the entire sample set cumulative species richness was calculated from 100 replicate random draws of each of (1, 12, 23, 34, ..., 920) 1-[m.sup.2] samples, drawn without replacement from a total of 920 samples.
The similarity of species composition and species densities between forest fragments and undisturbed continuous forest was measured using a quantitative similarity index, Normalized Expected Species Shared (NESS) (Grassle and Smith 1976, Wolda 1983). Similarity to undisturbed continuous forest was calculated as the mean similarity of each fragment site to the 14 continuous forest sites. Within-continuous-forest-site similarity was estimated from the average of all pairwise comparisons of species composition among the 14 continuous forest sites (n = 91).
To test whether the beetle fauna of fragmented forest sites was composed of widespread, disturbance-loving species, species similarity among replicate disturbed-area sites was contrasted with that among replicate sites in undisturbed forest (i.e., fragmentation should reduce beta diversity). For the purposes of analysis "disturbed-area sites" were defined as all 0-m and 13-m edge sites and both 1-ha sites (n = 10 sites).
Multivariate analyses.--Variation in species composition between sites was analyzed with Two Way Indicator Species Analysis, using the TWINSPAN computer program (Hill 1979b), and with Detrended Correspondence Analysis (DCA) and Canonical Correspondence Analysis (CCA) using the CANOCO (version 3.1) program (Hill 1979a, Gauch 1982, Ter Braak 1988, Jongman et al. 1995). Due to limitations on the number of species analyzed by the software available, a reduced sample set of 407 species and 7728 individuals was used (i.e., excluding all rare species represented by just one or two individuals). Rare species have little impact on multivariate analyses of species composition.
TWINSPAN is a polythetic divisive method of classification used to cluster sites of similar species composition and identify indicator species characteristic of different sites or groups of sites. Species' abundances were incorporated using three arbitrary pseudospecies cut levels (species abundances of 0, 2, and 10) (see Jongman et al. 1995), and the dendrogram was created with four levels of subdivision.
CCA is a multivariate analysis technique relating community composition to known variation in the environment. It provides an integrated description of species-environment relationships by assuming the existence of a single set of underlying environmental gradients to which all species respond (Ter Braak 1986). CCA can be used in combination with DCA to infer whether the measured environmental variables can account for the major variation in the species data (Ter Braak 1986).
Beetle species composition was related to the 12 environmental gradients detailed above. In addition, an interaction term for LNDIST x TEMP was included in CCA analyses, because edge transects were known to have differing temperature profiles depending on the density of vegetation buffering the forest edge (see Table 1 and Didham and Lawton, in press). It was also suspected that edge effects may vary with area (LNAREA x LNDIST, and LNAREA x LNDIST x TEMP). CANOCO 3.1 uses a forward selection procedure to rank environmental variables in order of their importance for determining species composition, in much the same way as forward stepwise multiple regression (Ter Braak and Verdonschot 1995). The first variable selected is the variable with the highest marginal eigenvalue (fit when entered as the only environmental variable in the analysis). Subsequent variables entered are those with the highest conditional eigenvalues (additional fit, after adding previous variables), until no more variables explain significant variation in species composition. Significance at each step is tested by a Monte Carlo permutation test with 999 random permutations under the null model of no effect; if the observed multivariate partial F ratio is within the highest 5% of the F ratios for randomly generated data sets, the null hypothesis is rejected at the 5% significance level.
Partial CCAs were used to partial out the effect of one or more environmental variables by adding these into the analysis as covariables, while focusing on the remaining variables of particular interest (Jongman et al. 1995).
Species responses to environmental gradients were analyzed by tracing projections of species points onto the trajectory of the environmental arrow of interest; the order of the projection points indicates the approximate ranking of the centers of distributions of beetle species along the gradient. Rules for the interpretation of species-environment biplots follow Ter Braak (1986), Jongman et al. (1995) and Ter Braak and Verdonschot (1995).
Population densities.--Trends in beetle species population densities were analyzed with respect to fragment area, distance from forest edge, and measured environmental variables using backward, stepwise, multiple regression. Variables were removed from the model when P [is greater than] 0.05. To minimize spurious effects due to low sample size, only the 32 most abundant species (N [is greater than or equal to] 46 individuals, representing 3.2% of total species richness and 47% of total beetle abundance across all sites) were included in the analysis (Appendix C). Theoretically, if each common species were distributed at random there would be an equal probability of detecting species presence at all 46 sites. Significant variation in population density was accepted at a Bonferroni-corrected [alpha]' of 0.00156. Multivariate population density responses were used to develop an empirical classification of species responses to forest fragment area and distance from forest edge.
"Species loss" rates from forest fragments.--The number and percent loss of species from an undisturbed continuous forest pool of 29 abundant beetle species (i.e., 3 of the 32 most abundant beetle species were not found in continuous forest) were calculated for 1-ha, 10-ha, and 100-ha fragments. Measured "species losses" are of course absences from samples, and actually consist of two components: (1) genuine species loss (species absent); and (2) sampling error (species present, but undetected). As species become rarer, the probability of sampling errors (false zeros) increases.
For 100-ha fragments, sampling effort was identical to that in undisturbed continuous forest (280 [m.sup.2]), so a simple comparison of species lists was performed. For 1-ha and 10-ha fragments, sampling effort was only 40 [m.sup.2], so the number of species absent from samples may be biased by small sample size. Consequently, 50 independent, random draws of two sites (40 [m.sup.2]) were taken from the 14 continuous forest sites and these species lists compared with those from 1-ha and 10-ha fragments. The mean ([+ or -] 1 SE) percentage loss of species from these random lists is presented for 1-ha and 10-ha fragments.
Total number and percentage loss of species from samples were further broken down, albeit very roughly, into "directional" and "random" species losses: we inferred "directional species losses" (i.e., losses due to altered population dynamics) where significant density trends with decreasing fragment area or distance from edge were found, and "random species losses" (i.e., losses due to stochastic presence/absence at the time of fragment isolation) where species showed no significant trend in density with decreasing fragment area or distance from edge.
Species attributes conferring susceptibility to forest fragmentation.--We tested four attributes of species that may confer susceptibility to forest fragmentation: trophic group (as defined above), body size (mean body length of all individuals, in millimeters), rarity (mean density per square meter in undisturbed forest, total area sampled 280 [m.sup.2]), and population variability (coefficient of variation, CV, for density per square meter in undisturbed forest, %). The 32 common species were divided into three groups based on their population density responses to fragment area, edge effects, and environmental gradients: (1) species significantly negatively affected by fragmentation; (2) species not significantly affected by fragmentation; and (3) species significantly positively affected by fragmentation. Trophic group proportions, mean body size, mean population density, and population variability were compared between these three species groups.
Population density responses were also correlated with the likelihood of species loss with decreasing fragment area. For each of the 32 species the probability of absence from samples ("local extinction") was calculated for each fragment size, irrespective of distance from forest edge (1-ha and 10-ha, n = 2 sites; 100-ha, n = 14 sites; continuous forest, n = 28 sites), and the trends in these probabilities analyzed using logistic regression (Didham et al. 1998). For the purposes of analysis continuous forest was enumerated as 10 000 ha. With a Bonferroni correction for multiple tests, trends in probabilities of local extinction with decreasing fragment area were considered statistically significant at P [is less than] 0.00156. A full analysis of the data is presented in Didham et al. (1998). Here we compare P values for trends in local extinction probabilities between the three species groupings, and relate these to differences in species attributes.
The beetle fauna
A total of 8454 beetles comprising 993 species were sampled. Beetles represented 11% of total invertebrate abundance, with an average ground density of 7.83 beetles/[m.sup.2] in undisturbed continuous forest (total area sampled 280 [m.sup.2]). Dominant beetle families in the assemblage were Staphylinidae (particularly the subfamilies Pselaphinae, Aleocharinae, Paederinae, and Osoriinae), Scydmaenidae, Ptiliidae, and Curculionidae (particularly Scolytinae and Curculioninae), with 50-150 species each (Appendix A). Samples were taken from an extremely large species pool, as evidenced by the number of species represented by a single individual in samples (45%), and the steepness of species accumulation curves (Fig. 1). The accumulation curve for undisturbed continuous forest was lower than that for the entire sample set, a good indication that there was a significant change in species composition between continuous and fragmented forests.
[Figure 1 ILLUSTRATION OMITTED]
Species richness and composition
Absolute species richness increased significantly towards the forest edge, and showed nonlinear changes with decreasing fragment area (polynomial regression, all parameters entered regression at P [is less than] 0.002, [r.sup.2] = 0.52, Fig. 2A). A linear model of changes in absolute species richness with fragment area was significant ([F.sub.2,43] = 9.48, P [is less than] 0.0005, [r.sup.2] = 0.27), but explained less variation in the data. Variation in species richness was largely due to changes in beetle density (correlation between species richness and sample abundance, r = 0.91, P [is less than] 0.0001, Fig. 2B), and rarefied species richness did not vary significantly across sites ([F.sub.2,43] = 3.16, NS, Fig. 2C).
[Figure 2 ILLUSTRATION OMITTED]
Although rarefied species richness remained constant following fragmentation, species composition compared with undisturbed forest changed significantly (logistic regression on percentage similarity, G = 40.66, df = 2, P [is less than] 0.0001, Fig. 2D); both independently with decreasing distance from forest edge (LNDIST [adjusted for LNAREA]: G = 18.08, df = 1, P [is less than] 0.0001); and independently with decreasing fragment area (LNAREA [adjusted for LNDIST]: G = 20.01, df = 1, P [is less than] 0.0001). Background species similarity among sites within undisturbed continuous forest was 39.6% (raw data; or 44.9%, fitted surface), due to random sampling effects (Fig. 2D).
Edge sites (particularly at 0 and 13 m) and 1-ha sites were the least similar to continuous forest (Fig. 2D), but there was no indication that highly disturbed sites shared a common beetle fauna. Mean ([+ or -] 1 SE) species similarity among the 10 most disturbed sites (29.0 [+ or -] 1.6%, n = 45 pairwise comparisons) was not significantly higher than the mean similarity between disturbed sites and continuous forest (t test, t = 1.33, df = 53, NS); edge sites were as different from each other as they were from undisturbed forest sites. Surprisingly, mean species similarity among disturbed sites was also significantly lower than mean species similarity among undisturbed forest sites (t = -6.45, df = 134, P [is less than] 0.0001); i.e., fragmentation increased beta diversity.
TWINSPAN multivariate analysis indicated that overall changes in species composition were largely due to a dichotomy between fragmented and nonfragmented forest (Fig. 3). Within these groupings there was also a lesser degree of separation based on distance from forest edge or fragment area. Most of the important indicator species for major dichotomies were among the most abundant beetle species (see Appendix C) and thus have robust predictive power. However, fragment area was significantly correlated with ground distance east to west (r = -0.80, P [is less than] 0.0001). Thus TWINSPAN site clustering may also result from spatial autocorrelation of species composition at varying scales: sites along the same transect appeared to cluster more closely than sites between transects, and the rank ordering of sites in TWINSPAN was significantly correlated with ground distance east to west (Spearman's R = 0.77, P [is less than] 0.001) (Fig. 3). Clustering broadly followed the 34-km east-to-west separation of sites, but note that: (1) three undisturbed continuous forest sites clustered within the forest fragments grouping, despite wide site separation in space (20 km); and (2) fragment sites that clustered together were sometimes [is greater than] 10 km apart (Fig. 3).
[Figure 3 ILLUSTRATION OMITTED]
Relationship of species composition to environmental gradients
In a CCA analysis, GRNDDIST (ground distance from east to west in the study area), LNDIST (distance from forest edge), LNAREA (fragment area), CANHEIGHT (canopy height), TEMP (air temperature), LITBIOM (litter biomass), %TWIG (percent ground cover of twigs), LNDIST x TEMP (distance from edge x air temperature interaction), and LITMOIST (litter moisture content) explained significant variation in beetle species composition (Table 2, Fig. 4, Monte Carlo permutation test, 999 random permutations, P = 0.001). CCA eigenvalues for the first four axes (0.34, 0.32, 0.22, and 0.20, respectively) were not markedly lower than eigenvalues for DCA (0.41, 0.30, 0.23, and 0.17, axes 1-4, respectively), and all species-environment correlations were high (0.97, 0.95, 0.94, and 0.93, CCA axes 1-4, respectively), indicating that the measured environmental variables explained the major variation in species composition across sites. CCA eigenvalues and the lengths of the vectors fitted to environmental variables (or "arrows" for environmental variables in the terminology of Ter Braak 1986) (Fig. 4) indicate the importance of environmental gradients in explaining species composition. The environmental arrows in CCA axes 1 and 2 ordination space (Fig. 4) account for 40% of the variance in the weighted averages of beetle species with respect to the nine environmental variables. From Table 3 we infer that the first axis separated sites based on fragment area and ground distance east to west within the study area, with large-area sites at the eastern end of the BDFFP having a tall forest canopy and low percentage ground cover of twigs. The second axis represents an edge effect gradient, along which sites and species were ordered according to distance from forest edge, air temperature, and litter biomass.
[Figure 4 ILLUSTRATION OMITTED]
TABLE 2. Determining environmental variables that account for significant variation in beetle species composition between sites by their marginal (left) and conditional (right) effects on beetle species, as obtained by the CCA forward selection procedure. Marginal effects (forward: step 1) j Variable [[Lambda].sub.1] P 1 GRNDDIST 0.29 0.001 2 LNAREA 0.28 0.001 3 LNDIST 0.27 0.001 4 CANHEIGHT 0.27 0.001 5 LITBIOM 0.26 0.001 6 %TWIG 0.26 0.014 7 TEMP 0.26 0.001 8 %LEAF 0.25 0.001 9 LITDEPTH 0.22 0.022 10 CANDENS 0.20 0.012 11 LITMOIST 0.18 0.051 12 LNDIST X TEMP 0.17 0.052 Conditional effects (forward: continued) j Variable [[Lambda].sub.a] P cum ([[Lambda]. sub.a]) 1 GRNDDIST 0.29 0.001 0.29 2 LNDIST 0.26 0.001 0.55 3 LNAREA 0.23 0.001 0.78 4 CANHEIGHT 0.22 0.001 1.00 5 TEMP 0.17 0.001 1.17 6 LITBIOM 0.17 0.006 1.34 7 %TWIG 0.15 0.025 1.49 8 LNDIST X TEMP 0.16 0.013 1.65 9 LITMOIST 0.15 0.018 1.80 Note: [[Lambda].sub.1] = eigenvalue (fit) with variable j only; [[Lambda].sub.a] = increase in eigenvalue (additional fit); cum [[Lambda].sub.a] = cumulative total of eigenvalues [[Lambda].sub.a]; P = significance level of effect, as obtained with a Monte Carlo permutation test under the null model with 999 random permutations. Prior to forward selection, LNAREA X LNDIST, LNAREA X LNDIST X TEMP, and EVAP were removed due to multicollinearity among environmental variables. TABLE 3. Beetle species composition data from Fig. 4: canonical coefficients and the intraset correlations of environmental variables with the first four axes of CCA. Canonical coefficients Axis variable 1 2 3 4 GRNDDIST 0.404 -1.619 -1.503 0.363 LNDIST -0.143 -0.610 -0.786 1.475 LNAREA -0.220 -0.914 -1.568 -0.510 CANHEIGHT -0.411 0.284 0.212 0.046 TEMP -0.129 0.510 -0.614 2.075 LITBIOM -0.099 0.688 -0.114 -0.114 %TWIG 0.348 -0.065 0.028 -0.568 LNDIST x TEMP 0.084 -0.014 0.630 -0.933 LITMOIST 0.152 0.178 0.181 0.384 Correlation coefficients Axis variable 1 2 3 4 GRNDDIST 0.841 -0.131 0.064 -0.130 LNDIST -0.592 -0.512 0.109 0.429 LNAREA -0.768 0.101 -0.432 -0.271 CANHEIGHT -0.728 -0.128 0.343 -0.333 TEMP 0.418 0.626 -0.191 0.123 LITBIOM 0.547 0.514 -0.415 -0.153 %TWIG 0.680 0.320 -0.291 -0.047 LNDIST x TEMP 0.138 0.107 0.362 -0.139 LITMOIST -0.212 -0.257 0.267 -0.356
Ground distance east to west explained significant variation in species composition (Table 2, Fig. 4), as suspected from the spatial autocorrelation of species composition observed in the TWINSPAN analysis (Fig. 3). To quantify the sole effect of ground distance east to west (i.e., excluding confounding effects of fragment area), a partial CCA was carried out, with all significant environmental variables from Table 2 as covariables, and ground distance east to west (as a measure of beta diversity) as the variable of interest. Site ordering along axis 1 ([[Lambda].sub.1] = 0.23) was subjected to a Monte Carlo permutation test (999 random permutations) and ground distance was found to explain significant variation in species composition over and above variation due to fragmentation effects or measured environmental heterogeneity (P [is less than] 0.001) (Fig. 5). The magnitude of the effect was small relative to the influence of all covariables (sum of eigenvalues for all covariables [[lambda].sub.sum] = 1.57), but of the same order of magnitude as some significant individual environmental gradients, such as distance from forest edge or fragment area (see Table 2).
[Figure 5 ILLUSTRATION OMITTED]
To investigate the specific effects of fragmentation, ground distance east-west within the study area was partialled out of further analyses. A partial CCA with ground distance as a covariable removed most of the putative fragment area effect (although it was still significant) (Fig. 6). Following forward selection, eigenvalues and species-environment correlations for all four axes were marginally lower in the partial CeA analysis (eigenvalues: 0.32, 0.23, 0.22, and 0.20; correlation coefficients: 0.94, 0.94, 0.97, and 0.93; for axes 1-4, respectively), and overall explanatory power decreased from 30.8 to 28.3%. Interpretation of the axes is unambiguous (Table 4): the first axis is defined by distance from forest edge, air temperature, and litter biomass; the second axis by canopy height and percent twig cover; and the third axis by fragment area. Eigenvalues show that the extracted gradients are long.
[Figure 6 ILLUSTRATION OMITTED]
TABLE 4. Beetle species composition data from Fig. 6, with ground distance (as a measure of gamma diversity) partialled out: canonical coefficients and the intraset correlations of environmental variables with the first four axes of CCA. Canonical coefficients Axis variable 1 2 3 4 LNAREA -0.892 0.615 -1.499 -0.413 LNDIST -0.823 0.363 -1.005 1.168 TEMP 0.180 -0.078 -1.238 1.838 LNDIST x TEMP 0.121 -0.191 0.822 -0.810 CANHEIGHT 0.130 -0.819 -0.476 0.089 %TWIG 0.016 0.478 0.420 -0.793 LITBIOM 0.630 -0.166 -0.286 -0.166 Correlation coefficients Axis variable 1 2 3 4 LNAREA -0.018 0.028 -0.773 -0.555 LNDIST -0.784 0.025 0.230 0.388 TEMP 0.736 0.108 -0.204 0.240 LNDIST x TEMP 0.132 -0.500 0.069 0.024 CANHEIGHT -0.264 -0.872 -0.180 -0.228 %TWIG 0.458 0.636 0.051 -0.164 LITBIOM 0.711 0.313 -0.409 -0.116
Species responses to environmental gradients
Edge effects.--Species responses to edge effects were determined from a partial CCA with ground distance as a covariable (Fig. 6). The centers of distributions of the 32 abundant beetle species ordered strongly along the first axis in response to distance from forest edge, with little scatter along the second axis (Fig. 6). Strongly edge-avoiding species included Stelidota sp. 0569 (Nitidulidae), ?Agathidium sp. 0422 and Aglyptinus sp. 0584 (Leiodidae, Leiodinae), Phaenostoma sp. 0307 (Hydrophilidae, Sphaeridiinae), Piestus sp. 0575 (Staphylinidae, Piestinae), and Araptus sp. 0924 (Curculionidae, Scolytinae). Edge specialists included Goniacerus sp. 0724, Phalepsoides sp. 0770, Globa sp. 0720 and Jubus sp. 0793 (Staphylinidae, Pselaphinae), ?Ptenidium sp. 0606 (Ptiliidae), Araptus sp. 0887 (Curculionidae, Scolytinae), sp. 0179 (Staphylinidae, Aleocharinae), Phaenostoma sp. 0306, Tachys sp. 0269 (Carabidae, Trechinae), and Baeocera sp. 0541 (Staphylinidae, Scaphidiinae). Notably, only 1 of the 6 edge-avoiding species listed is a predator, while 7 of the 10 edge specialists listed are predators (see Appendix C for trophic group assignments). Despite this tendency, however, there was no statistically significant difference in guild proportions between edge-specialists (to the right of the origin in Fig. 6) and edge-avoiders (to the left of the origin in Fig. 6) (G test of independence, G = 4.31, df = 3, NS).
In multiple regression analyses, 15 of the 32 species (47%) showed significant trends in population density with: (1) decreasing distance from forest edge; (2) environmental gradients; and/or (3) decreasing fragment area (Table 5). Distance from forest edge (LNDIST) accounted for much of the significant variation in population densities, presumably as a composite function of various microclimatic and environmental edge gradients. Air temperature, litter moisture content, evaporative drying rate, and canopy height all explained significant variation in population densities in combination with other edge and fragment area effects, although individual environmental variables were rarely the sole predictors of observed trends. One exception to this was Araptus sp. 0929, which responded solely to litter moisture content (Fig. 7A). The importance of combined environmental effects is most clearly seen for three species, sp. 0179, Tachys sp. 0269, and Phaenostoma sp. 0306, which prefer sites close to the forest edge (for ill-defined reasons), yet also show a secondary preference for cool, moist conditions not usually associated with edges (e.g., Fig. 7B, C, Table 5). Thus, they are actively selecting the more mesic sites available at edges.
[Figure 7 ILLUSTRATION OMITTED]
TABLE 5. Population density responses of beetle species to forest fragmentation. Significant trends in population densities were found for 15 of 32 abundant (N [is greater than or equal to] 46) beetle species tested with backward, stepwise multiple regression on distance from forest edge, fragment area, and nine measures of environmental heterogeneity (see Appendix B). Species code Family Genus 0002 Staphylinidae: Tachyporinae Coproporus sp. 0179 Staphylinidae: Aleocharinae ?genus 0268 Carabidae: Trechinae Tachys sp. 0269 Carabidae: Trechinae Tachys sp. 0306 Hydrophilidae: Sphaeridiinae Phaenostoma sp. 0307 Hydrophilidae: Sphaeridiinae Phaenostoma sp. 0313 Leiodidae: Catopinae Adelopsis sp. 0541 Staphylinidae: Scaphidiinae Baeocera sp. 0575 Staphylinidae: Piestinae Piestus sp. 0793 Staphylinidae: Pselaphinae Jubus sp. 0887 Curculionidae: Scolytinae Araptus sp. 0905 Curculionidae: Scolytinae Hypothenemus sp. 0919 Curculionidae: Scolytinae Araptus sp. 0929 Curculionidae: Scolytinae Araptus sp. 0980 Curculionidae: Curculioninae ?genus Species code Minimum adequate model F df 0002 LNAREA + TEMP 147.200 2, 44 0179 Intercept + LNDIST + EVAP 8.039 2, 43 0268 Intercept + LNAREA + LNDIST 13.907 2, 43 0269 LNAREA + LNDIST + CANHEIGHT + TEMP 51.003 4, 42 0306 Intercept + LITMOIST + TEMP 8.903 2, 43 0307 Intercept + LNAREA + LNDIST + CANHEIGHT 30.791 3, 42 0313 Intercept + LNAREA 27.441 1, 44 0541 LNAREA + LNDIST 69.900 2, 44 0575 LNDIST 37.166 1, 45 0793 LNAREA 47.238 1, 45 0887 Intercept + %TWIG + LITDEPTH 21.387 2, 43 0905 LNAREA 40.761 1, 45 0919 LNAREA 183.610 1, 45 0929 Intercept + LITMOIST 13.456 1, 44 0980 LNAREA 70.146 1, 45 Parameter estimates (1 SE) Species code P [r.sup.2] Intercept 1 2 0002 <0.0001 0.864 ... 0.239 -0.344 (0.014) (0.110) 0179 0.0011 0.238 1.464 -0.138 -2.635 (0.311) (0.037) (0.793) 0268 <0.0001 0.365 -1.070 0.146 0.148 (0.384) (0.052) (0.050) 0269 <0.0001 0.813 ... 0.201 -0.309 (0.025) (0.029) 0306 0.0006 0.260 -6.590 0.103 0.603 (1.858) (0.028) (0.151) 0307 <0.0001 0.665 2.492 -0.151 0.250 (0.360) (0.030) (0.029) 0313 <0.0001 0.370 -0.626 0.184 (0.268) (0.035) 0541 <0.0001 0.750 ... 0.169 -0.111 (0.018) (0.023) 0575 <0.0001 0.440 ... 0.101 (0.017) 0793 <0.0001 0.501 ... 0.115 (0.017) 0887 <0.0001 0.475 -1.754 0.096 0.051 (0.454) (0.018) (0.016) 0905 <0.0001 0.464 ... 0.092 (0.014) 0919 <0.0001 0.799 ... 0.242 0.018 0929 0.0010 0.217 -3.506 0.076 (1.351) 0.021 0980 <0.0001 0.60l ... 0.108 (0.013) Parameter estimates (1 SE) Species code 3 4 0002 0179 0268 0269 0.042 -0.414 (0.009) (0.087) 0306 0307 -.068 (0.016) 0313 0541 0575 0793 0887 0905 0919 0929 0980 Note: Variables were removed at P > 0.05. The minimum adequate model is the model after stepwise removal of nonsignificant parameters. Overall model significance was accepted at a conservative, Bonferroni-corrected [Alpha] of 0.00156.
Conversely, edge preference in the twig-and-leaf-petiole-boring Araptus sp. 0887 appeared to be mediated by resource availability, as evidenced by its significant response to variation in litter depth and percentage twig cover (Fig. 7D, Table 5).
Fragment area effects.--Species were weakly ordered along the LNAREA arrow on the third axis of a partial CCA (Fig. 8), with a small overall gradient length and wide lateral scatter (in response to the stronger gradient of edge effects). Species found predominantly in continuous forest included Goniacerus sp. 0724, ?Ptenidium sp. 0606, Tachys sp. 0269, Jubus sp. 0793, Hypothenemus sp. 0905 (Curculionidae, Scolytinae), sp. 0980 (Curculionidae, Curculioninae), and Phalepsoides sp. 0770. In contrast, species found more typically in small forest fragments included sp. 0179, Globa sp. 0720, Araptus sp. 0887, and Phaenostoma sp. 0306. Species to the right of the LNAREA arrow in Fig. 8 are edge specialists, while species to the left are edge avoiders, as shown in Fig. 6.
[Figure 8 ILLUSTRATION OMITTED]
In multiple regression analyses, fragment area had a dominant influence on variation in population densities between sites (Table 5). There were more species showing a significant preference for large-area sites than showing other types of responses. Interestingly, species found more commonly in small fragments (sp. 0179, Phaenostoma sp. 0306, and Araptus sp. 0887; see Fig. 8) appeared to be responding to environmental factors, rather than reduced fragment area per se (Table 5).
An empirical classification of species responses to fragmentation.--Species were empirically classified by significant population density responses (Table 5) into four major categories (Table 6): (a) edge sensitive, area insensitive; (b) area sensitive, edge insensitive; (c) edge and area sensitive; and (d) edge and area insensitive. Within these categories, trends in density were either positive (deep-forest species), or negative (disturbed-area species), with species showing the full spectrum of responses to fragmentation (Table 6). Two species were edge avoiders (type A-I response, e.g., Fig. 9A); three species were edge specialists (type A-II response, e.g., Fig. 9B); five species were large-area specialists (type B-I response, e.g., Fig. 9C); there were no small-area specialists (type B-II response); and there were one to two species in all type C categories except C-IV (i.e., no small-area, edge specialists) (e.g., Fig. 9D-F, Table 6). Seventeen species (53%) were not significantly affected by forest fragmentation (category D). Of the species showing significant trends in population density, the vast majority were adversely affected by fragmentation, having significantly lower densities near forest edges and/or in small forest fragments. Only one species showed significant small-area preference.
[Figure 9 ILLUSTRATION OMITTED]
TABLE 6. An empirical classification of beetle species responses to forest fragmentation. Code Family Genus A) Species responding to edge effects only (area insensitive) I) Edge avoiders 0575 Staphylinidae Piestus sp. 0929 Curculionidae Araptus sp. II) Edge specialists 0179 Staphylinidae ?genus 0306 Hydrophilidae Phaenostoma sp. 0887 Curculionidae Araptus sp. B) Species responding to area effects only (edge insensitive) I) Large-area specialists 0313 Leiodidae Adelopsis sp. 0793 Staphylinidae Jubus sp. 0905 Curculionidae Hypothenemus sp. 0919 Curculionidae Araptus sp. 0980 Curculionidae ?genus II) Small-area specialists (category empty) C) Species responding to both edge and area effects I) Large-area, edge avoiders 0002 Staphylinidae Coproporus sp. 0268 Carabidae Tachys sp. II) Small-area, edge avoiders 0307 Hydrophilidae Phaenostoma sp. III) Large-area, edge specialists 0269 Carabidae Tachys sp. 0541 Staphylinidae Baeocera sp. IV) Small-area, edge specialists (category empty) D) Species showing no response to edge or area effects 0018 Staphylinidae Carpelimus sp. 0312 Leiodidae Adelopsis sp. 0356 Dytiscidae Copelatus sp. 0422 Leiodidae ?Agathidium sp. 0569 Nitidulidae Stelidota sp. 0584 Leiodidae Aglyptinus sp. 0606 Ptiliidae ?Ptenidium sp. 0714 Scydmaenidae Euconnus sp. 0720 Staphylinidae Globa sp. 0724 Staphylinidae Goniacerus sp. 0770 Staphylinidae Phalepsoides sp. 0785 Staphylinidae Tuberoplectus sp. 0888 Curculionidae Araptus sp. 0918 Curculionidae Araptus sp. 0924 Curculionidae Araptus sp. 0985 Curculionidae ?genus 0986 Curculionidae ?genus Species absent from samples taken in forest fragments of size: Code 1 ha 10 ha 100 ha A) Species responding to edge effects only (area insensitive) I) Edge avoiders 0575 X X 0929 II) Edge specialists 0179 X 0306 0887 X B) Species responding to area effects only (edge insensitive) I) Large-area specialists 0313 X X X 0793 X X X 0905 X X 0919 X 0980 X X II) Small-area specialists (category empty) C) Species responding to both edge and area effects I) Large-area, edge avoiders 0002 X 0268 X X X II) Small-area, edge avoiders 0307 III) Large-area, edge specialists 0269 X X X 0541 X IV) Small-area, edge specialists (category empty) D) Species showing no response to edge or area effects 0018 0312 X 0356 X 0422 X X 0569 X 0584 0606 0714 0720 0724 X 0770 0785 0888 0918 0924 X 0985 0986 Note: Significant trends in population densities for species in categories A-C are quantified in Table 5, and examples are shown graphically in Fig. 9. Species absences from forest fragments indicate absence from all sites in that size category.
Species absences from samples taken in forest fragments
Table 6 shows species absences from samples taken in forest fragments of 1 ha, 10 ha, and 100 ha. Ignoring three species that were not found in undisturbed continuous forest (Tachys sp. 0269, Phaenostoma sp. 0306, and Araptus sp. 0887), 49.8% of common species were not detected in samples from 1-ha forest fragments (i.e., they were absent, or present but too rare to be sampled), 29.8% from 10-ha fragments, and 13.8% from 100-ha fragments (Table 7). Estimates of "directional" vs. "random" species absences were determined from analyses of population densities (Table 5): most species absences from samples taken in forest fragments were of species that showed significant (directional) trends in density with fragmentation; and, as expected, both directional and random species absences from samples increased with decreasing fragment area (Table 7).
TABLE 7. Species loss rates of beetles from samples taken in tropical forest fragments. The total number and percentage loss of species from an undisturbed continuous forest pool of 29 abundant beetle species are shown (i.e., 3 of the 32 most abundant beetle species were not found in samples from continuous forest), for 1-ha, 10-ha, and 100-ha fragments. Directional Random Fragment size No. % (1 SE) No. % (1 SE) 1 ha 9([dagger]) 38.3 (0.62) 5([dagger]) 11.5 (0.66) 10 ha 6([dagger]) 24.2 (0.57) 2([dagger]) 5.6 (0.48) 100 ha 3 10.3 1 3.4 Total Fragment size No. % (1 SE) 1 ha 14([dagger]) 49.8 (0.72) 10 ha 8([dagger]) 29.8 (0.82) 100 ha 4 13.8 Note: For 1-ha and 10-ha fragments, species losses calculated by comparisons of species lists with continuous forest ([dagger]) may be biased by unequal sample sizes. Consequently, mean ([+ or -] 1 SE) percentage loss of species is calculated by comparison with 50 randomly generated species lists drawn with equal sample size from undisturbed continuous forest. "Directional" species losses are inferred for species that showed significant trends in population density with decreasing fragment area or distance from edge, and "random" species losses for species that did not show significant trends in density with decreasing fragment area or distance from edge (Table 5).
Species attributes conferring susceptibility to extinction
As expected, the 10 species that showed significant decreases in population densities with decreasing fragment area or decreasing distance from forest edge were significantly more likely to be absent from small forest fragments than the 17 species that did not show significant trends in population densities with increasing degree of fragmentation (t test on P values for logistic regression trends in local extinction probability with decreasing fragment area, t = -3.62, df = 25, P [is less than] 0.002; Table 8); i.e., declining density is a precursor to local extinction.
TABLE 8. Attributes of beetle species with varying population density responses to forest fragmentation: (1) negative density responses, (2) no significant density responses, or (3) positive density responses. For each species, P values are presented for logistic regression trends in the probability of absence from samples (local extinction) as a function of decreasing fragment area; the lower the P value, the more significant the increase in probability of absence from small forest fragments. P values are formally significant at a Bonferroni-corrected level of P < 0.00156. P value for trend in probability of local extinction with decreasing Species code Genus fragment area 1) Species negatively affected by fragmentation 0002 Coproporus sp. 0.001 0268 Tachys sp. [much less than] 0.001 0307 Phaenostoma sp. 0.002 0313 Adelopsis sp. [much less than] 0.001 0575 Piestus sp. 0.117 0793 Jubus sp. [much less than] 0.001 0905 Hypothenemus sp. 0.002 0919 Araptus sp. <0.001 0929 Araptus sp. 0.707 0980 ?genus [much less than] 0.001 2) Species not significantly affected by fragmentation 0018 Carpelimus sp. 0.270 0312 Adelopsis sp. 1.000 0356 Copelatus sp. 0.641 0422 ?Agathidium sp. 0.073 0569 Stelidota sp. 0.922 0584 Aglyptinus sp. 0.502 0606 ?Ptenidium sp. 0.058 0714 Euconnus sp. 0.485 0720 Globa sp. 0.856 0724 Goniacerus sp. 0.972 0770 Phalepsoides sp. 0.888 0785 Tuberoplectus sp. 0.535 0888 Araptus sp. 0.485 0918 Araptus sp. 0.709 0924 Araptus sp. 0.634 0985 ?genus 0.057 0986 ?genus 0.305 3) Species positively affected by fragmentation 0179 ?genus 0.838 0269 Tachys sp. [much less than] 0.001([dagger]) 0306 Phaenostoma sp. 0.002 0541 Baeocera sp. 0.004 0887 Araptus sp. 0.270 Mean Mean density body in continuous Population Trophic size forest variability, Species code group (mm) (individuals/ cv (%) [m.sup.2]) 1) Species negatively affected by fragmentation 0002 Pr 2.62 0.557 247 0268 Pr 1.97 0.471 234 0307 Pr 3.53 0.293 194 0313 S 1.64 0.086 419 0575 S 3.99 0.125 354 0793 Pr 1.46 0.075 475 0905 X 1.12 0.039 755 0919 X 1.27 0.346 329 0929 X 1.43 0.182 293 0980 X 1.85 0.096 331 2) Species not significantly affected by fragmentation 0018 S 1.85 0.129 376 0312 S 1.45 0.039 582 0356 Pr 3.87 0.039 582 0422 F 1.09 0.343 396 0569 S 1.89 0.007 1181 0584 F 1.17 0.104 433 0606 F 0.82 0.111 373 0714 Pr 1.20 0.132 357 0720 Pr 1.05 0.018 1673 0724 Pr 1.45 0.011 963 0770 Pr 1.03 0.018 743 0785 Pr 1.20 0.121 429 0888 X 1.16 0.093 454 0918 X 1.30 0.168 309 0924 X 2.08 0.143 1389 0985 X 2.96 0.186 278 0986 X 4.80 0.061 394 3) Species positively affected by fragmentation 0179 Pr 2.15 0.004 1673 0269 Pr 1.16 0.000 ... 0306 Pr 2.25 0.000 ... 0541 F 1.09 0.039 582 0887 X 1.17 0.000 ... Note: Trophic group assignments: F = fungivore, Pr = predator, S = saprophage, and X = xylophage. Body size is the mean of all individuals. Mean population density and population variability are calculated for 280 [m.sup.2] of leaf-litter samples from undisturbed continuous forest. ([dagger]) P values indicating a significant decrease in probability of local extinction from forest fragments, i.e., for species positively affected by fragmentation.
Trophic group proportions did not vary significantly between species negatively affected, species unaffected, and species positively affected by fragmentation (G test of independence, G = 2.97, df = 6, NS) (Table 8). Similarly, there was no significant variation in mean body size among species showing different responses to fragmentation (ANOVA, [F.sub.2, 29] = 0.52, NS) (Table 8).
In contrast, mean population density and population variability in undisturbed forest varied significantly between species in different response groups (Table 8). Intrinsically, density and population variability were strongly interrelated (Pearson's correlation coefficient, r = -0.51, n = 29, P [is less than] 0.005); i.e., small populations are more variable. Species that were negatively affected by fragmentation were significantly more abundant in undisturbed forest than species that were unaffected by fragmentation, or than species that were positively affected by fragmentation (ANOVA, [F.sub.2,29] = 6.40, P [is less than] 0.005). Furthermore, both density and population variability were strongly correlated with the probability of local extinction in small fragments. Across all species, population density was negatively correlated with P values for the probability of local extinction with decreasing fragment area (r = -0.54, n = 29, P [is less than] 0.003), and population variability was positively correlated (r = 0.64, n = 29, P [is less than] 0.0001); i.e., rare species with small, variable populations are significantly more likely to persist in small fragments.
Fragmentation and beetle species composition
Beetle species composition changed significantly with decreasing distance from forest edge and decreasing fragment area. Surprisingly, despite the change in species composition, rarefied species richness was roughly invariant across sites. In this case, estimates of "total biodiversity loss" are clearly meaningless, because rarefied species richness is not only dependent on loss rates of deep-forest species, but also on species replacement rates, presumably of gap specialists and other disturbance-adapted species (rates that are identical in this study). Curiously, however, we did not find that edge assemblages were composed solely of widespread, weedy species, or that all edges shared a common fauna. On the contrary, beetle species composition was more variable among edge sites than among undisturbed forest sites; thus, fragmentation appeared to increase beta diversity. At least part of the reason for this was the greater spatial separation of disturbed sites (undisturbed sites were in much closer proximity to each other). Nevertheless, species composition was undoubtedly highly variable at edges. For this to be the case, samples must have been drawn from an extremely large pool of disturbance-adapted species. Species establishment at edges, then, may have a strong stochastic component, making every tropical forest edge unique.
The principal dichotomy in a TWINSPAN classification of beetle species composition was between non-isolated, continuous forest sites and isolated, forest fragment sites. Secondary separation within the two clusters was strongly environmental, with sites at 026 m from the forest edge clustering distinctly from sites deeper within the forest. In addition, 1-ha fragments clustered with sites at 0 m from the forest edge, rather than with other sites at 52 m from the edge. Species composition in very small fragments was more different from continuous forest than expected from simple edge effects alone, because of area reduction and multiple edge influences (see Malcolm 1994).
A CCA analysis identified ground distance east to west as another important variable explaining variation in beetle species composition. Other significant variables in the CCA analysis were fragment area, distance from forest edge, and a number of environmental variables that co-varied with distance from edge (Didham and Lawton, in press): canopy height, air temperature, litter biomass, percentage ground cover of twigs, and litter moisture content. In addition, LNDIST x TEMP (a distance from edge x air temperature interaction effect) explained significant variation in the data, confirming the importance of physical edge structure in modifying microclimatic gradients along edge-to-interior transects (Didham and Lawton, in press). Beetle species composition responded strongly to these differing gradients in edge microclimate. The nine environmental variables explained the major variation in species composition, related primarily to fragment area and ground distance east to west along the first axis of CCA, and edge effects along the second axis.
The significance of species turnover in the tropics
Fragment area was highly correlated with site location due to the establishment of continuous forest control sites at the eastern end of the reserve network, and forest fragments toward the western end, at the time of creation of the BDFFP. As a result, the effects of fragment area and ground distance east to west on site associations in TWINSPAN and CCA were confounded, as suggested by preliminary analyses of a smaller subset of the same data (Didham 1997b). Adjacent sites along the same transects formed relatively discrete clusters in TWINSPAN, but this may be a trivial result of the clumped distributions of species and small-scale variability in environmental parameters. More importantly, overall site separation within the study area broadly followed the main 34-km east-to-west axis, suggesting that intrinsic species turnover with distance may account for some of the variation in species composition, rather than fragment area per se.
How serious might the confounding effects of area and location be? Species turnover occurs at many scales, from local to regional, but is most obvious between different habitat types (beta diversity) and over large distances, or across dispersal barriers, within the same habitat type (gamma diversity) (Cody 1993). In this study, there were no apparent differences in geology, soil type, weather, vegetation, elevation, or latitude, and no barriers to dispersal, such as rivers or mountains, along the 34-km east-to-west axis (Ribeiro 1976, Chauvel 1983, Salati 1985, Lovejoy et al. 1986, Lovejoy and Bierregaard 1990, Prance 1990, Bierregaard et al. 1992, Camargo 1992, Rankin-de-Merona et al. 1992, Camargo and Kapos 1995), but it is possible that there was an underlying environmental gradient that we failed to detect. Gamma diversity, on the other hand, is related to historical speciation patterns and the sizes of species' geographic ranges (Gaston 1990, 1991, Cody 1993) and is unlikely to operate at this spatial scale.
Intriguingly, with all other measured fragmentation and environmental variables factored out, ground distance east to west within the study area explained significant variation in species composition. Moreover, the strength of the gradient was of equivalent magnitude to environmental gradients, such as distance from forest edge or air temperature. It is likely, however, that the observed trend is created by unexplained local variation in species composition within the 100-ha 1 fragment (solid diamonds in Fig. 5), rather than by gradual species turnover with distance. Excluding 100-ha 1, partial CCA axis 1 scores show no trend with distance, because the two 1-ha fragments immediately adjacent to 100-ha 1 are not markedly different in "residual" species composition from undisturbed continuous forest.
Beetle species responses to fragmentation
The 32 most abundant species ordered strongly along partial CCA gradients of distance from forest edge and fragment area. Different species showed different responses, ranging from edge avoiders to edge specialists, and large-area specialists to small-area specialists. A number of these species were important indicators of particular site associations in the TWINSPAN analysis. The population densities of 15 of the 32 abundant beetle species (47%) were significantly affected by forest fragmentation. Species were classified by response into four major categories (Table 6): (A) edge sensitive, area insensitive; (B) area sensitive, edge insensitive; (C) edge and area sensitive; and (D) edge and area insensitive. Within these categories, trends in density were either positive (deep-forest species), or negative (disturbed-area species) (Table 6). The vast majority of species were adversely affected by fragmentation, having significantly lower densities near forest edges and/ or in small forest fragments. Few disturbed-area species reached high population densities in the study area, which accounts for the apparent absence of large numbers of species that are positively affected by fragmentation. Most edge species were rare to moderately common, and, moreover, were often localized to one or just a few edges. It is not possible to differentiate meaningful trends in population densities from intrinsic random sample effects for these rare species.
While most population density responses are qualitatively similar to responses exhibited by invertebrate species in previous studies (e.g., Powell and Powell 1987, Bach 1988, Clopton and Gold 1993, Bedford and Usher 1994, Margules et al. 1994), some responses are less intuitive. For example, Phaenostoma sp. 0307 (Hydrophilidae, Sphaeridiinae) is a small-area, edge avoider, and Tachys sp. 0269 (Carabidae, Trechinae) and Baeocera sp. 0541 (Staphylinidae, Scaphidiinae) are large-area, edge specialists. Some of these counterintuitive species responses may be chance events generated by the vagaries of sampling, but others may have ready biological explanations; for example, large-area, edge specialists may be canopy gap specialists during the adult stage of their life cycles, but nevertheless still require large areas of forest at other stages.
Edge responses were not readily explainable in terms of simple responses to environmental edge gradients. "Distance from forest edge" (LNDIST), as a composite variable incorporating measured and unmeasured variation in edge-related phenomena, was a better predictor of variation in species' population densities than individual environmental variables. The complexity of edges (and of biotic communities) makes simple surrogate variables such as air temperature poor predictors of changes in species composition. In the absence of mechanistic (experimental) tests of species responses to such gradients, the results shed only limited light on the biological basis for species' edge responses (let alone area responses). Nonetheless, the observed species-environment relationships may offer interesting insights into the natural history of the organisms. For example: among the twig-and-leaf-petiole-boring Scolytinae (Curculionidae), Araptus sp. 0887 was predominantly distributed at forest edges and in 1-ha fragments, in response to increased resource availability (%TWIG and LITDEPTH). Another species, Araptus sp. 0929, was more common in undisturbed forest than at edges as a result of sensitivity to litter moisture conditions. Still other species, Araptus sp. 0919 and Hypothenemus sp. 0905, were only found in large patches of forest, with no discernible environmental explanation. And yet other species, Araptus sp. 0888, Araptus sp. 0918, and Araptus sp. 0924, were widely distributed and apparently insensitive to habitat modification. Such varied responses within a single genus (Araptus) are striking. Evidently, closely related species do not always show similar responses to fragmentation, perhaps because of strong partitioning of space and resources (among other factors) between related "trophic species." On the plus side, this suggests that there may be surprisingly little loss of generic diversity in fragmented habitats. On the minus side, there is no substitute for the analysis of patterns at the species level.
As another example, species showing strong preference for edges (e.g., sp. 0179, Tachys sp. 0269, and Phaenostoma sp. 0306) are not simply distributed at random at edges, but instead differentiate between microsites based on environmental conditions. Edge species are more abundant at selected cooler or moister sites near to the edge, suggesting that the more extreme edge conditions may be near the upper tolerance limits even for edge-loving species.
Life history traits are clearly important in the response of beetle species to fragmentation, but the data permit few generalizations at this time. Other species attributes, however, do appear to make some species more susceptible to fragmentation than others. Food web theory predicts that species at higher trophic levels are more susceptible to extinction in fragmented or otherwise disturbed habitats (Pimm and Lawton 1977, Pimm 1991, Lawton 1995, Holt 1996), but empirical evidence is conflicting (e.g., Brown 1978, Patterson 1984, Kareiva 1987, Mikkelson 1993, Kruess and Tscharntke 1994, Schoener et al. 1995, Didham et al. 1998). In this study, the CCA analysis indicated that edge-specialist species were predominantly predators, while edge-avoiding species were predominantly fungivores or saprophages, but the overall trend among just the 32 common species was not significant. For the full species assemblage, Didham et al. (1998) found that proportions of species in different trophic groups varied significantly with distance from forest edge; the forest edge had a significantly higher proportion of predator species and a significantly lower proportion of xylophages (but proportions of species of herbivores, fungivores, saprophages, and xylomycetophages did not vary). Predator species were more affected by forest fragmentation than species in lower trophic levels (Didham et al. 1998).
Mean body size had no influence on the type of population density response shown by individual species (i.e., positively, negatively, or not affected by fragmentation).
Rarity and population variability, on the other hand, did have a significant influence on the probability of local extinction (i.e., absence from samples) with decreasing fragment area. The two attributes are highly interdependent (small populations are more variable), and in practice there is no way of distinguishing the relative importance of their effects. In concert, the two factors are traditionally thought to make populations more vulnerable to extinction (Leigh 1981, Diamond et al. 1984, Pimm et al. 1988, Gaston 1994), but instead we found that common beetle species were significantly more likely to disappear from small forest fragments than rare species. Recent models of multispecies coexistence under disturbance suggest that competitively superior species (dominant, abundant species) are, on average, poorer dispersers than inferior competitors, and are also the most susceptible to habitat destruction (Tilman et al. 1994). Furthermore, this effect is predicted to be most severe in tropical forests, where superior competitors are typically rarer than their counterparts in temperate forests and thus undergo deterministic extinction at proportionally lower rates of habitat destruction (Tilman et al. 1994). The generality of the required trade-off between competitive and dispersal abilities is far from established, but our data suggest that species that are more abundant in undisturbed forest are indeed more susceptible to fragmentation, while rare species are better at persisting.
Species loss rates from tropical forest fragments
Reduction in fragment area and environmental edge effects appear to cause a significant loss of species in 1-ha, 10-ha, and 100-ha forest fragments. Such species losses are only really estimable from trends in the population densities of abundant species, because of the likelihood of spurious results due to low sample size for rarer species. Nevertheless, of 29 abundant beetle species found in undisturbed continuous forest (together accounting for 48% of total beetle abundance), 49.8% of species were absent from samples taken in 1-ha fragments, 29.8% from 10-ha fragments, and 13.8% from 100-ha fragments. The degree of apparent species loss from 100-ha fragments is perhaps most surprising given that these beetles range in size from just 0.8 to 4.8 mm, and small invertebrates are not traditionally considered to require extensive tracts of forest to maintain populations (see also Didham 1997b).
Species loss, as measured by disappearance from samples, increased abruptly with decreasing fragment area, as might be expected if there are critical thresholds of fragment size or percent remaining habitat in the landscape that determine population extinction (cf. Lamont and Klinkhamer 1993, Andren 1996, Bascompte and Sole 1996). Yet some abundant species do persist, even in the smallest 1-ha fragments, indicating that there is not a complete turnover of species between continuous forest and highly disturbed fragments. Whether small fragments can act as long-term species refuges (Shafer 1995, Turner and Corlett 1996) is more difficult to ascertain due to long species' lag-times to extinction (Tilman et al. 1994).
These figures require very cautious interpretation as estimates of the absolute rates of local extinction of beetle species in fragmented forests. The degree to which the results for just 32 common species can be extrapolated to the entire assemblage is open to question. On the one hand, declines in local abundance of commoner species may not equate to local extinction; a species may be present but remain undetected in samples. On the other hand, the analyses were performed using only the commonest species. Overall, rarer species will presumably be more prone to local extinction than common species due to both initial absence from fragments (a sampling effect) and increased probability of extinction due to low absolute population size (Gaston 1994, Andren 1996). The relationship between rarity and local extinction is extremely complicated, however, because less common species appear to be better at persisting in fragmented habitats (possibly because they are better dispersers), yet at the extremes of rarity there must be a vanishingly small probability of establishment and survival in small forest fragments. For moderately common to common species, then, species loss rates appear to be proportional to population density due to a trade-off between dispersal and competitive abilities (Nee and May 1992, Tilman et al. 1994), but we expect that below some threshold population density absolute rarity outweighs dispersal advantages and species loss rates should be inversely proportional to population density.
Estimates of beetle species loss rates, as measured by absence from samples taken in forest fragments, were exceptionally high. Even at the centers of forest fragments as large as 100 ha there were significant changes in beetle species composition, including the apparent loss of 14% of the most abundant beetle species. Species loss rates for very rare species may well be markedly higher. The majority of the species analyzed were adversely affected by forest fragmentation, although the whole spectrum of individual species responses was shown from edge specialists to edge avoiders, and large-area specialists to small-area specialists. Variable species responses were responsible for the surprising finding of no change in rarefied species richness from the least disturbed to the most disturbed sites. In this system, there was no apparent loss of biodiversity because high species loss rates were balanced almost exactly by high species replacement rates, presumably of species from disturbed habitats.
Classification of species responses was a useful tool for predicting species absences from samples taken in forest fragments. Almost 50% of the species analyzed showed significant changes in population densities with forest fragmentation. These species were much more likely to be absent from forest fragments than species that showed random fluctuations in population densities, although some of the latter species were also absent from fragments, presumably by chance. Assuming the number of forest fragments in the landscape is reasonably high, and total habitat loss relatively low, most species that are simply random components of fragment assemblages should be able to persist in at least some of the fragments. The total loss of biodiversity will then approach the proportion of species that are absent from fragments due to directional changes in population densities: 10.3% of species from 100-ha fragments, 24.2% from 10-ha fragments, and 38.3% from 1-ha fragments. Surprisingly, the commonest species, and in particular those in higher trophic levels (Didham et al. 1998), appear to be the most susceptible to local extinction. The differential loss of these species will have a disproportionately greater effect on ecosystem process rates than if species were lost evenly across all trophic groups and abundance classes (Tilman et al. 1994, Didham et al. 1998). As the percentage of remaining habitat in the landscape decreases, biodiversity loss will increase due to the increasing importance of stochastic species losses. Recent simulation models suggest that losses will be even greater than expected from absolute rates of habitat loss, due to the existence of critical thresholds for population extinction (Andren 1994, 1996, Bascompte and Sole 1996).
Mandy Tocher, Stuart Hine, Claude Gascon, and field technicians at the BDFFP provided assistance and logistical support in Brazil. Invaluable taxonomic advice and help with species sorting was provided by Stuart Hine (Dytiscidae), Martin Brendell (various taxa), Roger Booth (q), Larry Kirkendall (Curculionidae, Scolytinae), Roger Beaver (Scolytinae), Chris Lyal (Curculionidae), Richard Thompson (Curculionidae, Cossoninae), Don Chandler (Staphylinidae, Pselaphinae), Michael Hansen (Hydrophilidae), Malcolm Kerley (Scarabaeidae), D. G. Halstead (Silvanidae), and M. S. Caterino (Histeridae). Funding for this study was provided by the Commonwealth Scholarship Commission and The British Council, UK; The Natural History Museum, London; the Smithsonian Institution, Washington, D.C.; INPA, Manaus, Brazil; the NERC Centre for Population Biology, Silwood Park, UK; and the University of Canterbury, New Zealand. This is publication number 186 in the BDFFP technical series.
Aizen, M. A., and P. Feinsinger. 1994. Habitat fragmentation, native insect pollinators, and feral honey bees in Argentine "Chaco Serrano." Ecological Applications 4:378-392.
Andren, H. 1994. Effects of habitat fragmentation on birds and mammals in landscapes with different proportions of suitable habitat: a review. Oikos 71:355-366.
--. 1996. Population responses to habitat fragmentation: statistical power and the random sample hypothesis. Oikos 76:235-242.
Bach, C. E. 1988. Effects of host plant patch size on herbivore density: patterns. Ecology 69:1090-1102.
Barbault, R., and S. Sastrapradja. 1995. Generation, maintenance and loss of biodiversity. Pages 192-274 in V. H. Heywood and R. T. Watson, editors. Global biodiversity assessment. United Nations Environmental Programme. Cambridge University Press, Cambridge, UK.
Bascompte, J., and R. V. Sole 1996. Habitat fragmentation and extinction thresholds in spatially explicit models. Journal of Animal Ecology 65:465-473.
Becker, P., J. S. Moure, and F. J. A. Peralta. 1991. More about Euglossine bees in Amazonian forest fragments. Biotropica 23:586-591.
Bedford, S. E., and M. B. Usher. 1994. Distribution of arthropod species across the margins of farm woodlands. Agriculture, Ecosystems and Environment 48:295-305.
Bellinger, R. G., F. W. Ravlin, and M. L. McManus. 1989. Forest edge effects and their influence on Gypsy Moth (Lepidoptera: Lymantriidae) egg mass distribution. Environmental Entomology 18:840-843.
Besuchet, C., D. H. Burckhardt, and I. Lobl. 1987. The "Winkler/Moczarski" eclector as an efficient extractor for fungus and litter Coleoptera. Coleopterist's Bulletin 41: 392-394.
Bierregaard, R. O., Jr., T. E. Lovejoy, V. Kapos, A. A. Santos, and R. W. Hutchings. 1992. The biological dynamics of tropical forest fragments. A prospective comparison of fragments and continuous forest. BioScience 42(11):859-866.
Bierregaard, R. O., Jr., and P. C. Stouffer. 1997. Understory birds and dynamic habitat mosaics in Amazonian rainforests. Pages 138-155 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago, Illinois, USA.
Brown, J. H. 1978. The theory of insular biogeography and the distribution of boreal birds and mammals. Great Basin Naturalist Memoirs 2:209-227.
Brown, K. S., Jr., and R. W. Hutchings. 1997. Disturbance, fragmentation, and the dynamics of diversity in Amazonian forest butterflies. Pages 91-110 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago, Illinois, USA.
Burbidge, A. A., and N. L. McKenzie. 1989. Patterns in the modern decline of Western Australia's vertebrate fauna: causes and conservation implications. Biological Conservation 50:143-198.
Camargo, J. L. C. 1992. Variation in soil moisture and air vapour pressure deficit relative to tropical rain forest edges near Manaus, Brazil. Thesis. Darwin College, University of Cambridge, Cambridge, UK.
Camargo, J. L. C., and V. Kapos. 1995. Complex edge effects on soil moisture and microclimate in central Amazonian forest. Journal of Tropical Ecology 11:205-221.
Cappuccino, N., and R. B. Root. 1992. The significance of host patch edges to the colonization and development of Corythucha marmorata (Hemiptera: Tingidae). Ecological Entomology 17:109-113.
Chauvel, A. 1983. Os latossolos amarelos, alicos, argilosos dentro dos ecossistemas das bacias experimentais do INPA e da regiao vizinha. Acta Amazonica Supplement 12:47-60.
Chauvel, A., M. Grimaldi, and D. Tessier. 1991. Changes in soil pore-space distribution following deforestation and revegetation: an example from the Central Amazon Basin, Brazil. Forest Ecology and Management 38:259-271.
Clopton, R. E., and R. E. Gold. 1993. Distribution and seasonal and diurnal activity patterns of Eutrombicula alfreddugesi (Acari: Trombiculidae) in a forest edge ecosystem. Journal of Medical Entomology 30:47-53.
Cody, M. L. 1993. Bird diversity components within and between habitats in Australia. Pages 147-158 in R. E. Ricklefs and D. Schluter, editors. Species diversity in ecological communities: historical and geographical perspectives. University of Chicago Press, Chicago, Illinois, USA.
Cusson, M., R. S. Vernon, and B. D. Roitberg. 1990. A sequential sampling plan for adult tuber flea beetles (Epitrix tuberis Gentner): dealing with "edge effects." Canadian Entomologist 122:537-546.
Diamond, J.M. 1984. "Normal" extinctions of isolated populations. Pages 191-246 in M. H. Nitecki, editor. Extinctions. University of Chicago Press, Chicago, Illinois, USA.
Didham, R. K. 1997a. An overview of invertebrate responses to forest fragmentation. Pages 303-320 in A. Watt, N. E. Stork, and M. Hunter, editors. Forests and insects. Chapman and Hall, London, UK.
--. 1997b. The influence of edge effects and forest fragmentation on leaf litter invertebrates in Central Amazonia. Pages 55-70 in W. F. Laurance and R. O. Bierregaard, Jr., editors. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago, Illinois, USA.
Didham, R. K., J. Ghazoul, N. E. Stork, and A. J. Davis. 1996. Insects in fragmented forests: a functional approach. Trends in Ecology and Evolution 11:255-260.
Didham, R. K., and J. H. Lawton. In press. Edge structure determines the magnitude of changes in microclimate and vegetation structure in tropical forest fragments. Biotropica.
Didham, R. K., J. H. Lawton, P. M. Hammond, and P. Eggleton. 1998. Trophic structure stability and extinction dynamics of beetles (Coleoptera) in tropical forest fragments. Philosophical Transactions of the Royal Society of London, Series B 353:437-451.
Doak, D. E, P. C. Marino, and P. M. Kareiva. 1992. Spatial scale mediates the influence of habitat fragmentation on dispersal success: implications for conservation. Theoretical Population Biology 41:315-336.
Fearnside, P. M. 1989. Deforestation and agricultural development in Brazilian Amazon. Interciencia 14:291-297.
--. 1990. Fire in the tropical rain forest of the Amazon Basin. Pages 106-116 in J. G. Goldammer, editor. Fire in the tropical biota. Springer-Verlag, New York, New York, USA.
Gaston, K. J. 1990. Patterns in the geographical ranges of species. Biological Review 65:105-129.
--. 1991. How large is a species' geographic range? Oikos 61:434-438.
--. 1994. Rarity. Chapman and Hall, London, UK.
Gauch, H. G., Jr. 1982. Multivariate analysis in community ecology. Cambridge University Press, Cambridge, UK.
Grassle, J. E, and W. Smith. 1976. A similarity measure sensitive to the contribution of rare species and its use in investigation of variation in marine benthic communities. Oecologia 25:13-22.
Hammond, P. M. 1990. Insect abundance and diversity in the Dumoga-Bone National Park, N. Sulawesi, with special reference to the beetle fauna of lowland rain forest in the Toraut region. Pages 197-254 in W. J. Knight and J. D. Holloway, editors. Insects and the rain forests of South East Asia (Wallacea). Royal Entomological Society of London, London, UK.
--. 1994. Practical approaches to the estimation of the extent of biodiversity in speciose groups. Philosophical Transactions of the Royal Society of London, Series B 345: 119-136.
Harper, L. H. 1989. The persistence of ant-following birds in small Amazonian forest fragments. Acta Amazonica 19: 249-263.
Hill, M. O. 1979a. DECORANA--a FORTRAN program for detrended correspondence analysis and reciprocal averaging. Cornell University, Ithaca, New York, USA.
--. 1979b. TWINSPAN--a FORTRAN program for arranging multivariate data in an ordered two-way table by classification of the individuals and attributes. Ecology and Systematics, Cornell University, Ithaca, New York, USA.
Holt, R. D. 1996. Food webs in space: an island biogeographic perspective. Pages 313-323 in G. A. Polis and K. O. Winemiller, editors. Food webs: integration of patterns and dynamics. Chapman and Hall, New York, New York, USA.
Hubbell, S. P., and R. B. Foster. 1988. Canopy gaps and the dynamics of a neotropical forest. Pages 77-96 in M. J. Crawley, editor. Plant ecology. Blackwell Scientific, Oxford, UK.
Jongman, R. H. G., C. J. F. Ter Braak, and O. F. R. Van Tongeren. 1995. Data analysis in community and landscape ecology. Cambridge University Press, Cambridge, UK.
Jordan, C. F. 1986. Local effects of tropical deforestation. Pages 410-426 in T C. Whitmore and J. A. Sayer, editors. Tropical deforestation and species extinction. Chapman and Hall, London, UK.
Kareiva, P. 1987. Habitat fragmentation and the stability of predator-prey interactions. Nature 326:388-390.
--. 1990. Population dynamics in spatially complex environments: theory and data. Philosophical Transactions of the Royal Society of London, Series B 330:175-190.
Kitahara, M., and K. Fujii. 1994. Biodiversity and community structure of temperate butterfly species within a gradient of human disturbance: an analysis based on the concept of generalist vs. specialist strategies. Researches on Population Ecology 36:187-199.
Klein, B. C. 1989. Effects of forest fragmentation on dung and carrion beetle communities in central Amazonia. Ecology 70:1715-1725.
Kruess, A., and T Tscharntke. 1994. Habitat fragmentation, species loss, and biological control. Science 264:1581-1584.
Lamont, B. B., and P. G. L. Klinkhamer. 1993. Population size and viability. Nature 362:211.
Laurance, W. F. 1991. Ecological correlates of extinction proneness in Australian tropical rain forest mammals. Conservation Biology 5:79-89.
Laurance, W. F., and R. O. Bierregaard, Jr., editors. 1997. Tropical forest remnants: ecology, management and conservation of fragmented communities. University of Chicago Press, Chicago, Illinois, USA.
Laurance, W. F., and E. Yensen. 1991. Predicting the impacts of edge effects in fragmented habitats. Biological Conservation 55:77-92.
Lawrence, J. F., and A. F. Newton, Jr. 1995. Families and subfamilies of Coleoptera (with selected genera, notes, references and data on family-group names). Pages 779-1006 in J. Pakaluk and S. A. Slipinski, editors. Biology, phylogeny, and classification of Coleoptera: papers celebrating the 80th birthday of Roy A. Crowson. Muzeum I Instytut Zoologii PAN, Warsaw, Poland.
Lawton, J. H. 1995. Population dynamic principles. Pages 147-163 in J. H. Lawton and R. M. May, editors. Extinction rates. Oxford University Press, Oxford, UK.
Lawton, J. H., and R. M. May, editors. 1995. Extinction rates. Oxford University Press, Oxford, UK.
Leck, C. F. 1979. Avian extinctions in an isolated tropical wet-forest preserve, Ecuador. Auk 96:343-352.
Leigh, E. G. 1981. The average lifetime of a population in a varying environment. Journal of Theoretical Biology 90: 213-239.
Lord, J. M., and D. A. Norton. 1990. Scale and the spatial concept of fragmentation. Conservation Biology 4:197-202.
Lovejoy, T. E., and R. O. Bierregaard. 1990. Central Amazonian forests and the Minimum Critical Size of Ecosystems Project. Pages 60-71 in A. H. Gentry, editor. Four neotropical rainforests. Yale University Press, New Haven, Connecticut, USA.
Lovejoy, T. E., R. O. Bierregaard, A. B. Rylands, J. R. Malcolm, C. E. Quintela, L. H. Harper, K. S. Brown, A. H. Powell, G. V. N. Powell, H. O. R. Schubart, and M. B. Hays. 1986. Edge and other effects of isolation on Amazon forest fragments. Pages 257-285 in M. E. Soule, editor. Conservation biology: the science of scarcity and diversity. Sinauer Associates, Sunderland, Massachusetts, USA.
Main, B. Y. 1987. Persistence of invertebrates in small areas: case studies of trapdoor spiders in Western Australia. Pages 29-39 in D. A. Saunders, G. W. Arnold, A. A. Burbidge, and A. J. M. Hopkins, editors. Nature conservation: the role of remnants of native vegetation. Surrey Beatty and Sons in association with CSIRO and CALM, Canberra, Australia.
Malcolm, J. R. 1994. Edge effects in central Amazonian forest fragments. Ecology 75:2438-2445.
Margules, C. R., G. A. Milkovits, and G. T. Smith. 1994. Contrasting effects of habitat fragmentation on the scorpion Cercophonius squama and an amphipod. Ecology 75:2033-2042.
Maurer, B. A., H. A. Ford, and E. H. Rapoport. 1991. Extinction rate, body size, and avifaunal diversity. Proceedings of the International Ornithological Congress XX:826-834.
McCann, J. M., and D. M. Harman. 1990. Influence of the intrastand position of black locust trees on attack rate of the locust borer (Coleoptera: Cerambycidae). Annals of the Entomological Society of America 83:705-711.
Mikkelson, G. M. 1993. How do food webs fall apart? A study of changes in trophic structure during relaxation on habitat fragments. Oikos 67:539-547.
Murcia, C. 1995. Edge effects in fragmented forests: implications for conservation. Trends in Ecology and Evolution 10:58-62.
Myers, N. 1986. Tropical deforestation and a mega-extinction spasm. Pages 394-409 in M. E. Soule, editor. Conservation biology: the science of scarcity and diversity. Sinauer Associates, Sunderland, Massachusetts, USA.
Nadkarni, N. M., and J. T. Longino. 1990. Invertebrates in canopy and ground organic matter in a Neotropical montane forest, Costa Rica. Biotropica 22:286-289.
Nee, S., and R. M. May. 1992. Dynamics of metapopulations: habitat destruction and competitive coexistence. Journal of Animal Ecology 61:37-40.
Nepstad, D. C., C. Uhl, C. A. Pereira, and J. M. E. Silva. 1996. A comparative study of tree establishment in abandoned pasture and mature forest in eastern Amazonia. Oikos 76:25-39.
Nepstad, D. C., I. C. Vieira, P. Moutinho, and P. Jipp. 1993. Forest fragments and the future flora and fauna of eastern Amazonia. Pages 5-7 in J. K. Doyle and J. Schelhas, editors. Forest remnants in the tropical landscape: benefits and policy implications. Smithsonian Migratory Bird Center, Smithsonian Institution, Washington, D.C., USA.
Patterson, B. D. 1984. Mammalian extinctions and biogeography in the Southern Rocky Mountains. Pages 247-293 in M. H. Nitecki, editor. Extinctions. University of Chicago Press, Chicago, Illinois, USA.
Pimm, S. L. 1991. The balance of nature? University of Chicago Press, Chicago, Illinois, USA.
Pimm, S. L., H. L. Jones, and J. Diamond. 1988. On the risk of extinction. American Naturalist 132:757-785.
Pimm, S. L., and J. H. Lawton. 1977. Number of trophic levels in ecological communities. Nature 268:329-331.
Powell, A. H., and G. V. N. Powell. 1987. Population dynamics of male euglossine bees in Amazonian forest fragments. Biotropica 19:176-179.
Prance, G. T. 1990. The floristic composition of the forest of Central Amazonian Brazil. Pages 112-140 in A. H. Gentry, editor. Four neotropical rain forests. Yale University Press, New Haven, Connecticut, USA.
Rankin-de-Merona, J. M., G. T. Prance, R. W. Hutchings, M. E da Silva, W. A. Rodrigues, and M. E. Uehling. 1992. Preliminary results of a large-scale tree inventory of upland rain forest in the Central Amazon. Acta Amazonica 22: 493-534.
Ribeiro, M. N. G. 1976. Aspectos climatologicos de Manaus. Acta Amazonica 6:229-233.
Robinson, G. R., R. D. Holt, M. S. Gaines, S. P. Hamburg, M. L. Johnson, H. S. Fitch, and E. A. Martinko. 1992. Diverse and contrasting effects of habitat fragmentation. Science 257:524-527.
Roland, J. 1993. Large-scale forest fragmentation increases the duration of tent caterpillar outbreak. Oecologia 93:25-30.
Roland, J., P. Taylor, and B. Cooke. 1997. Forest structure and the spatial pattern of parasitoid attack. Pages 97-106 in A. D. Watt, N. E. Stork, and M. D. Hunter, editors. Forest and insects. 18th Symposium of the Royal Entomological Society of London. Chapman and Hall, London, UK.
Rosenzweig, M. L. 1995. Species diversity in space and time. Cambridge University Press, Cambridge, UK.
Salati, E. 1985. The climatology and hydrology of Amazonia. Pages 18-48 in G. T. Prance and T. E. Lovejoy, editors. Key environments: Amazonia. Pergamon, Oxford, UK.
Saunders, D. A., R. J. Hobbs, and C. R. Margules. 1991. Biological consequences of ecosystem fragmentation: a review. Conservation Biology 5:18-32.
Sayer, J. A., and T. C. Whitmore. 1991. Tropical moist forests: destruction and species extinction. Biological Conservation 55:199-213.
Schoener, T. W., D. A. Spiller, and L. W. Morrison. 1995. Variation in the hymenopteran parasitoid fraction on Bahamian islands. Acta Oecologica 16:103-121.
Shafer, C. L. 1995. Values and shortcomings of small reserves. BioScience 45:80-88.
Soule, M.E. 1983. What do we really know about extinction? Pages 111-124 in C. M. Schonewald-Cox, S. M. Chambers, B. MacBryde, and L. Thomas, editors. A reference for managing wild animal and plant populations. Benjamin/Cummings, Menlo Park, California, USA.
Sowig, P. 1989. Effects of flowering plant's patch size on species composition of pollinator communities, foraging strategies, and resource partitioning in bumblebees (Hymenoptera: Apidae). Oecologia 78:550-558.
Ter Braak, C. J. F. 1986. Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology 67:1167-1179.
--. 1988. CANOCO--a FORTRAN program for canonical community ordination. Microcomputer Power, Ithaca, New York, USA.
Ter Braak, C. J. F, and P. F. M. Verdonschot. 1995. Canonical correspondence analysis and related multivariate methods in aquatic ecology. Aquatic Sciences 57:1015-1621.
Tilman, D., R. M. May, C. L. Lehman, and M. A. Nowak. 1994. Habitat destruction and the extinction debt. Nature 371:65-66.
Tracy, C. R., and T. L. George. 1992. On the determinants of extinction. American Naturalist 139:102-122.
Turner, I. M. 1996. Species loss in fragments of tropical rain forest: a review of the evidence. Journal of Applied Ecology 33:200-219.
Turner, I. M., and R. T. Corlett. 1996. The conservation value of small, isolated fragments of lowland tropical rain forest. Trends in Ecology and Evolution 11:330-333.
Vasconcelos, H. L. 1988. Distribution of Atta (Hymenoptera-Formicidae) in "Terra-firme" rain forest of central Amazonia: density, species composition and preliminary results on effects of forest fragmentation. Acta Amazonica 18: 309-315.
Vitousek, P. M. 1994. Beyond global warming: ecology and global change. Ecology 75:1861-1876.
Whitmore, T. C., and J. A. Sayer. 1992. Deforestation and species extinction in tropical moist forests. Pages 1-14 in T C. Whitmore and J. A. Sayer, editors. Tropical deforestation and species eextinction. Chapman and Hall, London, UK.
Whittaker, R. H. 1970. Communities and ecosystems. Macmillan, London, UK.
Wilson, E. O. 1988. The current state of biological diversity. Pages 3-18 in E. O. Wilson and F. M. Peter, editors. Biodiversity. National Academic Press, Washington, D.C., USA.
Wolda, H. 1983. Diversity, diversity indices and tropical cockroaches. Oecologia 58:290-298.
APPENDIX A Beetle species assemblage sampled from 920 [m.sup.2] of leaf litter in Central Amazonia. Total num- Total ber number of of spe- indi- cies viduals col- col- Trophic Family: subfamily lected lected group Acanthoceridae 15 55 F Aderidae 1 1 F Anobiidae 2 3 X Anthicidae 1 3 S Biphyllidae 1 1 F Buprestidae 2 2 H Carabidae 33 377 Pr Cerylonidae 9 21 F Chrysomelidae 16 52 H Ciidae 22 107 F Coccinellidae 2 2 Pr Colydiidae 1 5 F Corylophidae 12 34 F Curculionidae: Brachycerinae 2 2 H Curculionidae: Cossoninae 4 8 X Curculionidae: Curculioninae 47 588 H, X Curculionidae: Dryophthorinae 5 12 H, X Curculionidae: Platypodinae 1 4 XF Curculionidae: Scolytinae 54 1618 X, XF Dermestidae 3 3 S Dytiscidae 1 56 Pr Elateridae 6 6 H, Pr, X Endomychidae: Merophysiinae 2 4 F Endomychidae (other) 16 97 F Erotylidae 6 6 F Histeridae 19 26 Pr Hydraenidae 1 6 H Hydrophilidae 14 254 Pr Laemophloeidae 3 5 F Languriidae 3 6 F Latridiidae 1 2 F Leiodidae: Catopinae 7 148 S Leiodidae: Leiodinae 8 336 F Limnichidae 2 5 S Melandryidae 1 1 F Nitidulidae 24 105 F, H, S Phalacridae 3 3 F, H Ptiliidae 61 202 F, S Ptilodactylidae 4 10 S Rhysodidae 1 9 F Salpingidae: Prostominiinae 1 1 F Scarabaeidae: Aphodiinae 2 29 S Scarabaeidae: Melolonthinae 4 9 H Scarabaeidae: Scarabaeinae 15 86 S Scydmaenidae 133 939 Pr Silvanidae 4 10 F Staphylinidae: Aleocharinae 105 416 F, Pr Staphylinidae: Euaesthetinae 7 15 Pr Staphylinidae: Megalopininae 2 9 Pr Staphylinidae: Osoriinae 43 249 S Staphylinidae: Oxytelinae 8 210 S Staphylinidae: Paederinae 45 246 Pr Staphylinidae: Piestinae 4 112 S Staphylinidae: Pselaphinae 109 1172 Pr Staphylinidae: Scaphidiinae 23 133 F Staphylinidae: Staphylininae 18 101 Pr Staphylinidae: Steninae 4 14 Pr Staphylinidae: Tachyporinae 11 359 Pr Tenebrionidae: Lagriinae 27 123 H, S Tenebrionidae (other) 12 36 F, S Total 993 8454 Note: Trophic group assignments (Hammond 1990): F = fungivore, H = herbivore; Pr = predator, S = saprophage, X = xylophage, XF = xylomycetophage (specialists on "ambrosia" fungus in wood). Nomenclature follows Lawrence and Newton (1995). APPENDIX B Measured environmental variables at 46 continuous and fragmented forest sites in Central Amazonia. Arbitrary LNA GRND site REA LNDIST DIST TEMP EVAP number (ha) (m) (km) ([degrees] C) (mL/h) Interior 1 1 9.210 9.210 0.420 -0.18 0.08 2 9.210 9.210 0.407 -0.27 0.09 3 9.210 9.210 0.394 -0.19 0.08 4 9.210 9.210 0.368 -0.15 0.10 5 9.210 9.210 0.315 -0.54 0.09 6 9.210 9.210 0.210 -0.42 0.08 7 9.210 9.210 0.000 -0.31 0.10 Interior 2 8 9.210 9.210 2.840 0.02 0.10 9 9.210 9.210 2.827 0.06 0.11 10 9.210 9.210 2.814 0.05 0.11 11 9.210 9.210 2.788 0.21 0.13 12 9.210 9.210 2.735 0.24 0.13 13 9.210 9.210 2.630 0.55 0.22 14 9.210 9.210 2.420 -0.20 0.10 Edge 1 15 9.210 0.000 12.760 0.80 0.11 16 9.210 2.565 12.747 0.28 0.08 17 9.210 3.258 12.734 0.22 0.07 18 9.210 3.951 12.708 -0.18 0.07 19 9.210 4.654 12.655 -0.08 0.05 20 9.210 5.347 12.550 0.05 0.07 21 9.210 6.040 12.340 0.76 0.17 Edge 2 22 9.210 0.000 12.760 2.55 0.55 23 9.210 2.565 12.747 1.68 0.22 24 9.210 3.258 12.734 1.09 0.21 25 9.210 3.951 12.708 0.75 0.11 26 9.210 4.654 12.655 0.46 0.12 27 9.210 5.347 12.550 0.13 0.09 28 9.210 6.040 12.340 1.43 0.24 100-ha 1 29 4.605 0.000 32.780 1.72 0.46 30 4.605 2.565 32.767 0.72 0.21 31 4.605 3.258 32.754 0.40 0.20 32 4.605 3.951 32.728 0.13 0.08 33 4.605 4.654 32.675 0.29 0.08 34 4.605 5.347 32.570 0.17 0.10 35 4.605 6.040 32.360 -0.07 0.11 100-ha 2 36 4.605 0.000 22.860 3.51 0.75 37 4.605 2.565 22.860 2.50 0.40 38 4.605 3.258 22.860 1.90 0.39 39 4.605 3.951 22.860 1.06 0.21 40 4.605 4.654 22.860 0.51 0.16 41 4.605 5.347 22.860 0.80 0.18 42 4.605 6.040 22.860 0.95 0.25 10-ha 1 43 2.303 4.654 21.860 0.40 0.12 10-ha 2 44 2.303 4.654 13.860 0.40 0.12 1-ha 1 45 0.000 3.951 32.880 0.59 0.14 1-ha 2 46 0.000 3.951 33.780 0.59 0.14 Arbitrary CAN CAN LIT LIT LIT site HEIGHT DENS DEPTH BIOM MOIST number (m) (0-3) (mm) (g/[m.sup.2]) (%) Interior 1 1 25.2 2.7 26.5 566.1 68.3 2 23.6 2.7 25.4 575.3 70.1 3 23.3 2.7 17.3 490.4 69.1 4 20.0 2.7 27.5 516.9 68.0 5 19.1 2.8 20.6 634.7 61.6 6 22.9 2.4 22.9 497.0 68.9 7 21.2 2.5 17.8 592.1 69.0 Interior 2 8 25.2 2.8 27.8 520.7 66.0 9 23.3 3.0 28.1 509.8 65.4 10 21.7 2.5 20.5 543.5 67.2 11 23.2 2.9 16.4 463.7 67.9 12 20.8 2.8 20.3 541.9 68.1 13 24.5 2.9 14.5 462.8 63.0 14 25.2 2.6 31.2 440.2 66.6 Edge 1 15 20.7 2.8 32.7 636.8 71.6 16 25.4 2.6 31.0 585.6 70.1 17 27.7 2.4 30.6 642.8 69.9 18 28.7 2.3 29.2 588.3 69.0 19 29.5 2.3 25.2 708.1 67.0 20 29.0 2.5 26.0 636.7 67.6 21 31.9 2.4 16.0 609.7 65.4 Edge 2 22 14.9 1.7 44.4 846.9 51.3 23 26.8 2.2 19.5 671.4 66.0 24 24.6 2.2 24.5 688.1 70.6 25 27.0 2.1 32.6 582.9 66.5 26 28.6 2.4 17.8 585.5 62.0 27 26.6 2.3 21.5 516.8 71.7 28 26.5 2.2 18.5 551.0 61.4 100-ha 1 29 10.1 2.6 31.1 799.3 66.5 30 13.4 2.8 35.0 743.3 67.4 31 15.1 2.7 28.5 789.0 65.2 32 15.7 2.7 22.5 602.1 66.5 33 26.8 2.8 21.9 641.7 61.2 34 25.1 2.8 11.8 687.3 63.5 35 26.9 3.0 23.8 667.0 61.4 100-ha 2 36 8.6 1.8 24.0 767.2 48.3 37 23.3 2.5 29.4 537.0 60.0 38 26.6 2.5 28.4 620.1 59.7 39 25.1 2.0 27.9 677.9 64.2 40 24.0 2.6 26.0 607.9 66.5 41 26.7 2.5 36.6 500.8 64.8 42 27.3 2.3 33.6 510.5 62.1 10-ha 1 43 28.1 2.6 14.8 462.6 64.0 10-ha 2 44 23.1 2.2 18.0 534.3 67.1 1-ha 1 45 24.0 2.6 32.6 619.7 59.6 1-ha 2 46 12.5 2.6 57.2 794.9 65.3 Arbitrary site % % number LEAF TWIG Interior 1 1 77.5 7.5 2 85.0 5.0 3 77.5 7.5 4 75.0 10.0 5 70.0 5.0 6 55.0 7.5 7 80.0 10.0 Interior 2 8 85.0 5.0 9 85.0 5.0 10 72.5 5.0 11 70.0 7.5 12 77.5 5.0 13 57.5 12.5 14 72.5 7.5 Edge 1 15 55.0 15.0 16 75.0 10.0 17 80.0 5.0 18 77.5 5.0 19 72.5 10.0 20 55.0 10.0 21 80.0 5.0 Edge 2 22 65.0 15.0 23 65.0 15.0 24 40.0 22.5 25 80.0 5.0 26 77.5 7.5 27 80.0 5.0 28 80.0 10.0 100-ha 1 29 50.0 20.0 30 60.0 15.0 31 75.0 5.0 32 50.0 17.5 33 65.0 10.0 34 50.0 15.0 35 72.5 10.0 100-ha 2 36 42.5 50.0 37 75.0 10.0 38 80.0 10.0 39 65.0 10.0 40 72.5 10.0 41 65.0 10.0 42 75.0 12.5 10-ha 1 43 85.0 5.0 10-ha 2 44 77.5 10.0 1-ha 1 45 52.5 10.0 1-ha 2 46 67.5 12.5 APPENDIX C Total abundances and trophic group assignments for the 32 most abundant beetle species sampled from 920 [m.sup.2] of leaf litter in Central Amazonia. Species code Family: subfamily Genus 0002 Staphylinidae: Tachyporinae Coproporus sp. 0018 Staphylinidae: Oxytelinae Carpelimus sp. 0179 Staphylinidae: Aleocharinae ?genus 0268 Carabidae: Trechinae Tachys sp. 0269 Carabidae: Trechinae Tachys sp. 0306 Hydrophilidae: Sphaeridiinae Phaenostoma sp. 0307 Hydrophilidae: Sphaeridiinae Phaenostoma sp. 0312 Leiodidae: Catopinae Adelopsis sp. 0313 Leiodidae: Catopinae Adelopsis sp. 0356 Dytiscidae Copelatus sp. 0422 Leiodidae: Leiodinae ?Agathidium sp. 0541 Staphylinidae: Scaphidiinae Baeocera sp. 0569 Nitidulidae Stelidota sp. 0575 Staphylinidae: Piestinae Piestus sp. 0584 Leiodidae: Leiodinae Aglyptinus sp. 0606 Ptiliidae ?Ptenidium sp. 0714 Scydmaenidae Euconnus sp. 0720 Staphylinidae: Pselaphinae Globa sp. 0724 Staphylinidae: Pselaphinae Goniacerus sp. 0770 Staphylinidae: Pselaphinae Phalepsoides sp. 0785 Staphylinidae: Pselaphinae Tuberoplectus sp. 0793 Staphylinidae: Pselaphinae Jubus sp. 0887 Curculionidae: Scolytinae Araptus sp. 0888 Curculionidae: Scolytinae Araptus sp. 0905 Curculionidae: Scolytinae Hypothenemus sp. 0918 Curculionidae: Scolytinae Araptus sp. 0919 Curculionidae: Scolytinae Araptus sp. 0924 Curculionidae: Scolytinae Araptus sp. 0929 Curculionidae: Scolytinae Araptus sp. 0980 Curculionidae: Curculioninae ?genus 0985 Curculionidae: Curculioninae ?genus 0986 Curculionidae: Curculioninae ?genus Species Trophic code group Abundance 0002 Pr 283 0018 S 114 0179 Pr 46 0268 Pr 170 0269 Pr 79 0306 Pr 59 0307 Pr 165 0312 S 47 0313 S 88 0356 Pr 56 0422 F 196 0541 F 59 0569 S 48 0575 S 66 0584 F 76 0606 F 82 0714 Pr 205 0720 Pr 142 0724 Pr 48 0770 Pr 53 0785 Pr 195 0793 Pr 125 0887 X 214 0888 X 123 0905 X 87 0918 X 183 0919 X 399 0924 X 75 0929 X 205 0980 X 84 0985 X 145 0986 X 59 Note: Trophic group codes are as in Appendix A. Voucher specimens are deposited at the Department of Entomology, INPA, Manaus, Brazil, and The Natural History Museum, London, UK.
RAPHAEL K. DIDHAM,(1,2,3,5) PETER M. HAMMOND,(1) JOHN H. LAWTON,(2) PAUL EGGLETON,(1) AND NIGEL E. STORK(1)
(1) Biodiversity Division, Department of Entomology, The Natural History Museum, Cromwell Road, London SW7 5BD UK (2) NERC Centre for Population Biology, Imperial College at Silwood Park, Ascot, Berks SL5 7PY UK (3) Projeto Dinamica Biologica de Fragmentos Florestais, INPA Ecologia/V-8, CP 478, 69011 Manaus, AM, Brasil (4) Cooperative Research Centre for Tropical Rainforest Ecology and Management, James Cook University, P.O. Box 6811, Cairns, Queensland 4870 Australia
Manuscript received 11 December 1996; revised 25 July 1997; accepted 4 August 1997; final version received 4 September 1997.
(5) Present address: Department of Biology, University of Delaware, Newark, Delaware 19716 USA.
|Printer friendly Cite/link Email Feedback|
|Author:||Didham, Raphael K.; Hammond, Peter M.; Lawton, John H.; Eggleton, Paul; Stork, Nigel E.|
|Date:||Aug 1, 1998|
|Previous Article:||Phosphorus dynamics in a lowland African rain forest: the influence of ectomycorrhizal trees.|
|Next Article:||Seed dispersal and the Holocene migration of woodland herbs.|